Inconsistent results using REW apparently ?

[Viabcroce]

Dear all,

first post here.
I've recently started to use REW in order to properly set my hi-fi room. I have been using it (latest Release version) in conjunction with a Umik-1 on an Asus PC 1101HA notebook (internal soundcard) running W7.

My audio system is based on an Anthem STR and a pair of B&W 706 S2 on SolildSteel stands.
Room is treated with 2 absorbing panels at the early refection points on the front wall, 3 at the early refection points on the ceiling + 2 mobile panels at the early refection points on the side walls.
The room is irregular (equilateral trapezoid), plus speakers and MLP are closer to one side (the 90° one) than the other.

I've first run a couple of identical SPL Measurements in order to check sw, hw and procedural consistency.
First of all, I've noticed REW doesn't allow me to run multiple sweeps, but one only ("No Timing Reference" is selected, "Number of Sweeps" greyed out).

Having said this, I ran three identical single Measurements. Here follow testing conditions:

• Umik calibrated for frontal positioning (not 90°).
• Tripod on the sofa, mic placed horizontally at the MLP aiming towards the centre of the two speakers.
• I leave the room while measurements run.

Here follow my observations and my doubts:

1. I'm getting some weird "comb filter-style" nulls in the frequency response. While I'd consider this to be normal in the low-freqs region, I'm getting them all over the spectrum (up to 18 kHz) (please refer to Forum differences.mdat).
   
   
2. The three measurements' plots should be, if not identical, at least quite superimposable. What I see instead is some large divergence (20-30 dB), for example @ 5945 Hz, 8040, 10210...
   
3. Interestingly enough, these big differences always occur at "negative spikes". There are other differences here and there but those are milder. Also, notice that spikes disappear between 11 and 13 kHz, appear again in the 13700-14200 kHz range and then become rarer (but don't disappear) at higher freqs (see for example 16300 Hz).
   I am pretty sure this means something, but can't say what.
   
   
4. While the sweeping frequency gets higher (around 17 kHz), I can almost always hear other sounds (apparently sub-harmonics) overlapping in the last seconds. This is weirdly reflected in the plot.
   Also, feel free to check Forum warbling.mdat: this set is an example of a measurement where the sub-harmonics did (measurement 2,from about 16 kHz on) and did not show (measurement 1). Please note the two measurements were taken in different moments and at different levels, enforcing the hypothesis that there's something "randomly" going in REW, in W7 or in the notebook.
   

I'm wondering whether all the above might hint at some sw or hw issues in the testing system (or even some wrong procedures from my part)?
I cannot test/measure my audio system, but from what I can say I can't perceive sonic issues when listening to music programs.

Thanks for your help everyone!

Gianfranco Di Mare

 

[John Mulcahy]

Multiple sweeps will not work when the input and output are different devices as their clocks will not be synchronised (as discussed in the help). When REW detects a USB mic it automatically disables multiple sweeps as the output must be a different device.

If you have more than one speaker playing then there will be comb filtering from their combined signal at the mic, most evident at high frequencies. Measure speakers individually.

Make sure the sample rate chosen for the default format in the Windows Audio properties for the output device (not sure what it is as you haven't made any selection in REW) is 48 kHz, the same as the REW sample rate, otherwise the OS will be resampling between the REW rate and the Windows audio default format rate.

Those 3 measurements are about as identical as it is possible to get.

 

[skid00]

I would also suggest that you measure at least a few feet *away* from the couch. That really messed up my measured response. Can you more the couch to the back wall?

I mount my UMIK in foam, then zip-tie that to a 90* tripod arm, to get away from reflections. That might be what you are seeing at 18K and up. I had that phenomenon pop up on some captures, and when I rotated the tripod's 3 feet to be as far away from the mic as possible, may have corrected those reflections.

I also found that using the Graph menu, Var Smoothing and 1/12 octave to be more representative of what I hear.

The Harman Room Calculator will help to position your speakers: https://www.harman.com/documents/Room Mode Calculator_0.xls

 

[Viabcroce]

Thanks guys for responding

Multiple sweeps will not work when the input and output are different devices as their clocks will not be synchronised (as discussed in the help). When REW detects a USB mic it automatically disables multiple sweeps as the output must be a different device.

This is interesting. Actually, what you specified is correct to the letter: in fact, if I use a mic that REW doesn't detect, multiple sweeps are allowed.
This is for example the case with an ARC mic which doesn't request a cal file: REW doesn't signal it has found a mic, but everything works nevertheless. Probably at a lower, OS level.
That was the first mic I used when I started becoming familiar with measurements, and I used to set 4 sweeps per measurement.

If you have more than one speaker playing then there will be comb filtering from their combined signal at the mic, most evident at high frequencies. Measure speakers individually.

The attached file Forum L e R soli.mdat shows L and R channels measured individually. I am not sure the number of shown destructive interactions is the same, but for sure there are a lot. Still wondering whether this is the real room, or some mistake I am making. I checked your suggestions about audio settings were respected.
Something that puzzles me in this regard is the following: the file Forum Destructive.mdat shows two measurements of the R channel. Everything was identical, except that in 2 I had put an absorbing mobile panel (GIK Gobo, i can't post links ATM) on the back of the couch, exactly centered on the MLP/mic placement.
I can't understand how an absorbing panel can create new destructive interferences, while (and this sounds more reasonable to me) cancelling some others. See for example 1725 and 1790 Hz, but they are all over the spectrum.

I would also suggest that you measure at least a few feet *away* from the couch. That really messed up my measured response.

Hey skid, yep, that makes sense. But I wanted to get a measurement that was more representative of how the room sounds in the MLP. I'd find odd if i treated the room according to what the mic measures far from where my head is?

Can you more the couch to the back wall?

The couch is in the middle of the room, but against a column. I can move it closer to the front wall, but not the other way round...
What were you thinking of?

I mount my UMIK in foam, then zip-tie that to a 90* tripod arm, to get away from reflections. That might be what you are seeing at 18K and up.

Yep, I use that kind of tripod as well. Also, I make sure the main pole is always as far as possible from the mic's vertical.
As for the rippling you mention, I hear(d) a second "melody"/subharmonic in that region when the sweep would go. It was not present when I got the measurements I am attaching now, and in fact the related graphics don't show it. It might have been to do with the sampling rate mismatch (44.1/48) I had on my measurement system?
Anyway, I'll try and see if I can reproduce it

I also found that using the Graph menu, Var Smoothing and 1/12 octave to be more representative of what I hear.

Yep, I've noticed that even 1/48 dramatically changes the curves' appearance. I don't have much experience yet, so I am never sure whether I am hiding something I should take into consideration or if that is only "floor noise" with respect to what measurement mean in terms of listening experience...

The Harman Room Calculator will help to position your speakers

Very kind from you, thanks
The problem in my case is that the room is irregular (a rectangular trapezium), is not symmetrical in terms of doors, windows, columns, plus i can't place speakers and MLP close to any median.
So my strategy ATM is to correct the room "mechanically" as I can (panels, ERP etc) and then recur to DRC.

Thanks again for you interest all

 

[John Mulcahy]

Every surface in the room generates reflections that contribute to comb filtering. Panels only absorb a proportion of the incident sound (that proportion will vary with frequency and the design and material of the panel), the rest is reflected.

 

[Viabcroce]

Every surface in the room generates reflections that contribute to comb filtering. Panels only absorb a proportion of the incident sound (that proportion will vary with frequency and the design and material of the panel), the rest is reflected.

Yep, I get that. But how can adding a panel generate tens of new cancellations all over the spectrum?
By reducing the amount of energy\frequencies going around, destructive interactions should be reduced.
I can understand that a solid frame can become a source for new reflections (25 dB tho?!), but the Gobo is totally covered, and double-face. And, only half of it faced above the couch's top

 

[John Mulcahy]

By adding a new source of reflections, large ones if it was placed close to the mic.

 

[Viabcroce]

By adding a new source of reflections, large ones if it was placed close to the mic.

Ah sorry, I had misread your comment
Yep, that makes sense. I'll be curious to compare the panel's absorption range/window with the distibution of the new nulls.

 

[skid00]

Hey skid, yep, that makes sense. But I wanted to get a measurement that was more representative of how the room sounds in the MLP. I'd find odd if i treated the room according to what the mic measures far from where my head is?

I am of the belief that our brains account for the response irregularities as we move about the couch. After all, you don't hear your TV, spouse, etc sounding odd when you sit there.

Anyway, create two EQ curves, one at the couch, one speaker nearfield, or couch moved back, and see what you prefer!

 

[Viabcroce]

I am of the belief that our brains account for the response irregularities as we move about the couch. After all, you don't hear your TV, spouse, etc sounding odd when you sit there.

No doubt sounds we hear changes according to thousands of factors, ad no doubt we can enjoy even classical music on a pocket radio. But this is not the point: hi-fi is about detection of issues and their solutions, and there's a setting/setpu at which your system sound closest to what was recorded. In hi-fi, we try and create reproducible conditions, or it would make no sense to choose a cable that sounds better in the low-mid region, or even placing loudspeakers at their best in relationship with the room, furniture and MLP.

Anyway, create two EQ curves, one at the couch, one speaker nearfield, or couch moved back, and see what you prefer!

What I "prefer" is to hear the sound with the least interferences. So if my head is usually close to the couch's back I am looking for improving my system's response at that location.
Of course I also listen to music from other locations in the room, but I like the idea of having an optimized spot, so to speak
As a side note, with the Anthem you can save up to 4 curves, and even account for the varying relative distance between speakers. In fact, I use a DRC curve for my MLP, one for when I lay on one side of the couch reading, one for those occasions where there's more than one person listening on the couch and a broaden sweet spot is desirable. Ah, I also have one for when I train on the carpet LOL

Generally speaking, this thread is not about how to set up a room or an audio system: it's about perspective errors I am making in using REW.