Working with REW

Wow! That looks a bit angry. Huge differences across the board. What is your listening room like? From the graphs I'd guess there's not much furniture in there, or at least the walls are bare? Im sure my room would have a similar looking graph. I must get more in there! 

That peak at 55hz hardly decays at all. I'm assuming it is a room resonance node. I'm surprised that you don't find it all a bit "boomy". Your mid-highs that would cause harshness don't look so bad, but there is a real roll-off from 6k that doesn't look good. That's perhaps a speaker issue, not a room issue. The S400s aren't super-bright and this is consistent with that, but they certainly shouldn't be harsh.

Thanks for the feedback, both.

Yes, it does look a bit angry.  The room is carpeted and furnished, but there's a big patio window down one side of it, which must be less than ideal in terms of reflections etc.  There are floor length curtains to the window, and I did another sweep with the curtains drawn.  But it wasn't substantially different, really.  Leather sofas and a coffee table probably don't help either.

Yep, there is some boominess at some bass frequencies, and the null at approx 40Hz is noticeable with some material too.

But none of the above gets in the way of enjoying the music.  As I (poorly) understand it, adding bass traps or Gik panels or similar could help with image coherence and that sort of thing.  And that'd be great, but it's not the immediate problem.  It's the hardness in the upper mids that grates on me.  To put it in excessively basic and exaggerated terms for illustration, I'd be more than happy to have a perfectly accurate acoustic image of a piano in my room, but in the first instance I'd be happy enough with a slightly blurred piano so long as the pianist's right hand wasn't poking me in the ear.  Likewise it's less important whether the bass player takes a step forward or back when the more pressing issue is the drummer jabbing you with his snare stick.  Like I say, those are exaggerated descriptions, but hopefully they convey the point.

Yet as Winky says, it looks fairly unproblematic in that region on the graphs.  I was kind of expecting there to be a spike or a significant ridge on the waterfall in that region, implying what might be called glare.

There are slight spikes in frequency response at approx 3, 4.6 and 6.2KHz, but they don't look particularly pronounced to me.  Having said that, I've read in some places that the ideal response at the listening position is a slight downward slope to the SPL graph - flat responses are desirable on speaker measurement charts in anechoic chambers, but not in listening rooms.  If true, that would imply that the whole 2-6KHz region is a bit louder than is ideal.

As it stands, although clearly the room isn't perfect, to my limited understanding there doesn't seem to be anything on the graphs which explains the issue I have.  So I'm thinking it's either an anomaly in my hearing or a black box/speaker issue.

Any other thoughts are more than welcome.

Have you tried tweeting the speaker positioning? Are they firing directly at you? Try having them facing straight down the room. Are they near any surfaces like side walls which may cause reflection issues? Or can your listing position be moved? 

I got in touch with GIK with pics and dimensions of my room. They were very helpful. With their advise I now have the system sounding the best it has in my room and any treatments I add, and I will, should improve things further. 

Yeah, I've moved the speakers about and eliminated toe in. A minor improvement (well, a tradeoff) but definitely not problem solved. I also tried moving the speakers into a totally impractical position just to see what happened when they were placed very differently in the room, and the issue was still there.  

The listening position can't really be moved other than a few inches back or forth, which I've also tried. It made some difference to the bass null, but didn't do much in the high mids.

And I've tried makeshift absorption at first reflection points. On the walls at least, I didn't try the ceiling. This improved the sound, but not in a way which affected the issue with the prominent high mids. That's why I'm skeptical about panels in the way I described in the pianist and drummer examples above. It's a bit like icing a cake that's got shards of glass in the sponge.

I can't help but feel that Gik or similar stuff would be premature at the minute because there's something more fundamental going on.

Dave***t posted:

I've been troubled by a perceived harshness in the high mids from my system for some time now.  

I've tried moving the speakers (a bit, at least, my room isn't big enough to allow radical moves), eliminating toe in, and I've tried using makeshift sound absorbers at various points on the walls (heavy towels).  I've gone back to stock everything in terms of cables, supports etc, I've done my best to ensure that no signal cables run too close or parallel to power cables.  Basically all of the obvious stuff to try.

All to no avail.

So I picked up a Umik-1 and downloaded REW, in order to try to get to the bottom of things.  And now I have a 20Hz-20KHz reading of the room taken from the listening position with a frequency response graph, a 'waterfall' plot and so on.

The thing is, I'm no expert at reading this kind of stuff.  There's some wrongness going on at the bass end.  But I don't have an immediate issue with the bass - I need to solve the high mid problem before worrying about that.  Likewise there's an unexpected peak around 18KHz which is way too high to be the issue.

It's perfectly possible that the problem stems from not getting on with the speakers (S-400) and/or the source (usually a 172, but for this test an offboard DAC running into the 172's analogue input) rather than a problem with the room.  Or from some other non-room related thing.  But this measurement seemed a good place to start.

I've attached images of the SPL graph and the waterfall - does anyone fancy helping me to interpret them?

you say you perceive harshness from some time now:  so before there was no perceived harshness ?   what changed in your system vs before ?

Have you tried changing the witch hat back to the naim din to xlr? And/Or replacing the powerline on the 172 for the standard mains cable? 

Monsieur Rooster, it sort of crept up on me, but I'd say it roughly coincided with getting the S400s and subsequently running them in (which notoriously takes a very long time). However I think that's a matter of them being revealing modern speakers and simply showing up what was always there. Hard to be sure tbh, bottom line is that it's there now, and doesn't match precisely to any individual component change.

And yes FF, I tried going back to stock everything, and then back again. Each thing made a difference to the sound, but not in the sense of making my issue disappear or get worse. Different differences, if that makes sense. Also tried swapping the speaker cables with some spare shorter lengths of A5 I have, same deal.

I'm going to try putting my old SBLs back in and running a sweep again after a couple of days' warm up. Fingers crossed...

Hi, well,

Firstly to address the bass (because its easiest to see the issue there) and the waterfall plot.

You have room peaks at 30Hz and ~57Hz, these are almost 1:2, so I'd guess your room is about 5.7m x 3m.  The 40Hz null isn't actually a null, it's just where the speakers aren't getting any room reinforcement.
The waterfall trace doesn't go on long enough (it's truncated at 300ms, to assess bass modes you need to extend that to about 800ms), also there's no point in having the minimum below 25dB (as below that just shows the general room noise clutter).

To look for the harshness you need to do three things

1  Smooth the FR and waterfall plots with 1/12th octave smoothing
2      Check this with 1/24th smoothing to see if any significant features appear
3  Post the raw impulse response plot over the first 5ms.  Do this separately for each speaker (disconnect at the amplifier end and run sweep), and with both combined.

Nice post Huge. Got to say that I am continually impressed by your knowledge, and by your willingness to help others on this forum. 

I think this graphic (from independentrecording.net) is useful to refer to when trying to match sounds perceived vs frequencies.

Hook posted:

Nice post Huge. Got to say that I am continually impressed by your knowledge, and by your willingness to help others on this forum. 

As am I, and I'm especially thankful in this case.

I checked out the smoothing etc, and it made a lot more sense of the graphs.  I couldn't collect the per speaker impulse response as I've swapped the SBLs back in as noted above (seemed worth a go, and I didn't really know what I was doing!).  So only the stereo information is available thus far.  If it's crucial I can either swap the speakers back again or perform another sweep with the SBLs (as it's a room measurement, one might suspect it was comparatively speaker agnostic).

The room size estimate is quite accurate.  It's 3.2m wide and 5.7m long, with a rectangular shaped bit nipped off the right rear corner where the door to the kitchen is, approx 2.1m long and 0.95m wide.  Speakers along the 3.2m wall, firing down the length of the room.  Due to window placement, domestic requirements etc, it has to be this way around, and firing across the width is not possible.

Hopefully more comprehensible plots:

1/12 smoothing SPL

1/24 smoothing SPL for comparison

Waterfall 1/12 smoothing & scale amended

And (potentially not very helpful, but since it was available...) stereo impulse response

Even if no one has any further comments or advice, at least I've learned how to make readable graphs with REW!

Your impulse response graph is out to 989ms, that makes it very hard to see early reflections in the 0-5ms range.

From the waterfall plot, it's clear that you do need bass traps (as shown by the very slow decays at 30, 57 and 90 Hz).

Apologies, I misread the desired scale.  Closer in on the first few ms:

Agreed that the bass end shows a need for bass traps.  But that depends to an extent on the outcome of the high mids issue.  

If that ends up with a speaker change, one of the options would be P3ESR, possibly plus a sub with Antimode DSP.  That'd change the bass situation considerably.  But if the high mid issue can be sorted without changing speakers, then I'll consider the bass problems afterwards.

Right, we have found the likely source of the harshness...

You have a number of early reflections at about...

0.7ms, this corresponds to a path difference of about 250mm
3.3ms, this corresponds to a path difference of about 1.1m
4.3ms, this corresponds to a path difference of about 1.5m
5.6ms, this corresponds to a path difference of about 1.9m

There may be more that aren't shown here.
Your room really needs diffusers at several critical reflection points.
The waterfall plot suggest that, overall, it's not underdamped, but having said that, some added MF / HF damping may well work to reduce the harshness.

Aha, fascinating, I see what's going on there.

Thanks again for the input, I'll have a play around later (ie try again with individual speakers, now that I understand the idea, and possibly try again with some makeshift materials on the walls/ceiling) and see what's what.

Yes, it's not as difficult to understand as some make out, is it!

Hi [@mention:41551092540348438] any updates? 

I shall be running REW at some point this week. I’m very interested in the results. I’m very much aware that my room needs some treatment, the walls are bare at the moment. I have a vaulted ceiling which is 3.1m high and I don’t know if that’s a good or bad thing. 

Hi.  Yes, after measuring, fiddling with calculators and measuring tapes (my other half thought I was mad, but bore it with admirable relaxation), sticking soft cushions at strategic points on walls and all of that sort of thing, I also tried some alternative gear.  And of all the things I tried, the most appreciable results came from the alternative gear.

So my conclusion is that the S400s are fine speakers, but I just don't get on with them in my room (with my possibly slightly quirky hearing).  So I'll be moving them on, and am just considering options at the moment.  Then once I've got their replacements here & sorted, I'll go back to REW for the fine tuning.

You'll have to let us know how you get on.

One interesting thing was that last weekend I borrowed a pair of little Harbeth speakers for comparison, and thought I'd take a sweep measurement with them while I was at it.  And despite being tiny, in my room they went far deeper in the bass than their spec says they're supposed to be able to.  Here's a plot with S400s (red), my previous SBLs (blue) which I pressed back into experimental service, and the P3ESRs (green).  Surprising result.

Fascinating stuff Dave. Those Harberths look like impressive little speakers. I’ll have to figure out a way of getting an image of my results up once I’ve run it this week 

Ok, so I had a brief moment on Sunday to try running the REW programme.  I connected everything up but sound only comes out of one channel. Is this correct? The odd thing is that the sound is from the right channel on my laptop, but left channel when connected to the 272. Same with headphones on the laptop, left channel only. All other sound works correctly on the laptop. I’ve looked through every setting on REW but can’t find any fix. 

Hi Finkfan,

In my case I use a USB->S/Pdif external "soundcard" and connect that to the 272, and I get sound from both channels.  I've not tried it with the analogue output of the laptop.

I used USB first, and then when I'd handed the USB DAC back to my mate, optical out from my iMac. In either case, both channels were definitely used, unless selected otherwise in REW.

So there's something going wrong there, Finkfan. Since it does the same thing both via line out (?) and headphones, I'd double check the REW settings and the laptop sound output settings I think.

Ok, thanks guys. I’ll check again tonight. Will check all settings I can find and maybe reinstall REW if that doesn't work. 

Stupid question, but worth asking just in case (I was caught by it for a bit once) - REW lets you choose whether the sound is left, right or both channels - it it is set in the 'measure' or 'generator' windows.