REW repeatability.

[Chrisbee]

Hi

I wondered about variability between tests when using REW.

I was seeing considerable differences between my curves particularly at the bottom end of the frequency range. I wondered what I could do to improve repeatability.

So just out of interest I kept repeating sets of 9 REW sweeps without changing anything.

Here is my IB calibrated to 75dB on the Galaxy 140 with 75dB target level in REW: Headroom average 25dB. REW warnings of low readings on each test.

Only 5-20Hz shown. Galaxy meter at listening position at ear height.



Below 10Hz the curves seem almost random with slight variations above.

Here I recalibrated to 80dB on the meter but without changing REW's 75dB target setting: Headroom average now 12.5dB.



Here I recalibrated the meter to 85dB leaving target level at 75dB. Headroom average now 9.2dB.



It seems if you want improved repeatability it pays to give REW enough signal to work with.

[JimP]

I wonder if this is background noise in the electronics?

What happens if you go lower with the SPL? does it scatter more?

[Chrisbee]

JimP wrote:

I wonder if this is background noise in the electronics?

What happens if you go lower with the SPL? does it scatter more?

I suppose I ought to try it to find out.

At these very low frequencies perhaps atmospheric pressure plays a part in the signal to noise ratio?

It is also quite windy here today and the windows are open "on the catch" for ventilation. The IB enclosure is not airtight either.

This may all help to explain the increased scattering below 10Hz with lower SPLs.

I thought it fascinating how the bundle of curves tightens with increased signal.

[JohnM]

You can also try using longer sweeps, which should similarly improve consistency and help in rejecting low frequency noise. Headroom would ideally be around 12dB for room measurements.

[Chrisbee]

Thankyou for these ideas, John.

I'll have play around and see what happens.

It's probably all quite academic on real programme material.

Though it's remarkable to be able to see such close results at these extremely low frequencies. Probably much of what the meter is reading is harmonic distortion.

My attempts to push 10Hz too high <95dB(C) produced a clearly audible higher harmonic. I need more displacement!

[JohnM]

One of the characteristics of the log sweep measurement technique is that harmonics are easily rejected, REW excludes them from the displayed frequency response. You can get a feel for how much the harmonic distortion content is increasing by looking at the portion of the impulse response that lies before the main peak (i.e. at times < 0), the harmonics appear as scaled down versions of the main impulse response. You can also use the percentage 2nd and 3rd harmonic displays on the SPL meter when playing sine wave test tones with the signal generator.

[Chrisbee]

JohnM wrote:

One of the characteristics of the log sweep measurement technique is that harmonics are easily rejected, REW excludes them from the displayed frequency response. You can get a feel for how much the harmonic distortion content is increasing by looking at the portion of the impulse response that lies before the main peak (i.e. at times < 0), the harmonics appear as scaled down versions of the main impulse response. You can also use the percentage 2nd and 3rd harmonic displays on the SPL meter when playing sine wave test tones with the signal generator.

Fascinating. Thankyou, John.

Here's my impulse response. There seems to be very little happening before the "0".



EDIT: Sorry, forgot about SPL meter under Tools in REW.

[Chrisbee]

Here are some interesting VLF curves: Sweep length now 1M. (Adjusted in details in Test box)
2-20Hz graph limits. 75dB REW target level. SPL meter reads 80dB(C)
Half the sweeps with and half without the C-weighting box ticked in meter settings.
The more horizontal curves below 6Hz are C-weighting. The Galaxy SPL meter cal file stops at 5Hz.
Amazing consistency above ~6.5Hz!
4 x 15" IB opposed driver manifold. Galaxy 140 meter, Fast, C-weighting, Mid SPL range.



I'll try the REW signal generator next for distortion measurements using the REW SPL function.

[Chrisbee]

REW Signal Generator using Frequency follows cursor function.

REW SPL function reading dB and 2nd harmonic distortion.

REW SPL function calibrated to 75dB.

Hz.....dB.....2nd%

50.....75......0.15
40.....75......0.2
30.....75......0.2
25.....75......1.2
20.....75......1.2
15.....74......0.5
10.....74......1.0
9.......69......1.5
8.......62......3.0
7.......60......5.5
6.......57......7.0
5.......40......90

Figures for 2nd harmonic distortion are rough averages.

3rd harmonic readings remained unchanged throughout.

These distortion figures look too low to my uneducated eye.

Perhaps I should up the levels?

[Chrisbee]

Front door slamming just as I hit "Measure". C-weighting ticked.