• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

REW Measurements - What does a good system measure like in the time / frequency domain?

OP
N

Naughtius

Member
Joined
Mar 4, 2022
Messages
56
Likes
70
First, you have a pretty big difference in the response of the left and right channels between 600 and 2000Hz which shows through in the impulse response as well.

That could have a pretty detrimental effect on stereo imaging, and I would try more symmetrical speaker placement within the room. If that is not possible, absorption for the right wall. Even something 2" thick would help, although 3.5" is a better bet.

You have some weird distortion peaks at 1000, 2000, 3000, and 4000Hz. Is there a problem with the measurement mic, or the amp powering the speakers?

All things considered, the response below 400Hz is pretty good, but there is still some elevated decay times. Nothing terrible, but I would put some 5.5" absorbers on the ceiling just to balance it out. Plus that will help with the nature of your speakers having lackluster vertical directivity but good horizontal directivity.

The entire treble region seems a bit elevated. I would consider pulling down 2-3dB above 1000Hz and evaluating if you like the change.


View attachment 190476


Adding a properly integrated sub would help with the bass deviations below 100Hz, particularly that 55Hz dip.

Thanks Alex, these are some nice pointers I can work on!

I'm not sure where the kHz peaks come from, they show up in the background measurement as well:
Background.jpg

I haven't heard them, so it's probably something with my UMIK.

I'll try playing with the house curves. What would you personally aim for on the decay times?
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,439
Likes
7,948
Location
Brussels, Belgium
Thanks! Which figures did you look at for this? And what values would one typically aim for here?

Any improvements over the big concrete box I'm in are probably due to my 110kg foam beanbag, here's a photo from when I just moved in: :D
View attachment 190610
The spectrogram, decay time is Y axis and frequency is X axis.

200 ms to 400 ms is where you want to be for music.
 

dc655321

Major Contributor
Joined
Mar 4, 2018
Messages
1,597
Likes
2,235
That's a good looking improvement! I activated a Dirac Live trial on my laptop and give it a go as well - Here's the measurements I'm getting compared to the IIR filters I was using. How would you interpret these differences?

View attachment 190611

You may want to have look at this thread, and @mitchco’s results for “ideal” time domain characteristics. Some good info in there.
 
OP
N

Naughtius

Member
Joined
Mar 4, 2022
Messages
56
Likes
70
'I'm still not quite happy with the sound as it is.'

Are you able to pin point what you are not happy with? Bass, vocals, imaging?

What does a good system measure like in the time / frequency domain?


Generally you want a smooth measuring response. For example the frequency response should not have sharp ups and downs but smoother undulations aiming towards flat, especially in the bass region - around 30-300Hz (bearing in mind your speakers probably go down to 30/40Hz). Overall I like flat but some like a slope with, say, 30Hz being 6dB higher then 10kHz (what are called 'house curves' and there are various preferences with these).

Actually, the best question is 'what does a good system sound like'. It's too easy to get caught up in measurements. It's what it sounds like to you that is the most important.

Generally it is best to solve bass problems first. These are the most difficult to deal with but when done allow the rest of the music to be revealed.

To solve bass issues involves up to four methods which can be use individually or together depending on your domestic situation and finance.

1. Positioning of speakers and listening chair (Ears).
2. Room treatment - bass traps etc..
3. Sub woofers - one or more carefully positioned.
4. DSP/EQ (e.g Dirac).

1. Cost's nothing but time and effort.
2. Costs something and may not be domestically acceptable.
3. Costs and takes time and effort to integrate.
4. Costs and takes time and effort to learn and understand. Dirac is about as easy as it gets.

I posted some stuff on my efforts here (I use 1, 2 and 4 - no subs) - see post 60 - which may help:

https://www.audiosciencereview.com/...in-room-measurements.13540/page-3#post-411614

Thanks for taking the time for this thorough reply!

My current setup sounds, well, the word that comes to mind is 'messy'. I visited my local HiFi store and listened to a random setup they had through two different amplifiers - a tube and solid state one - which, in comparison to my own setup, sounded very 'clean', even though I didn't like the coloring of the particular speakers. Here's the room I listened to:

Untitled.png


What comes to mind when looking at the differences;
- No coffee table in the sound path ;)
- Good thick rug
- No close side walls
- Nice dampening record collection right behind the speakers
- Very high ceiling

Your post on improving your sound in the other tread is also very helpful, and comparing the time domain measurements with mine shows a lot of room for improvement, also aiming mostly towards room treatment. Do you have any pictures of the final room? Curious to see!
 

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,239
Likes
1,384
Location
Budapest
That's a good looking improvement! I activated a Dirac Live trial on my laptop and give it a go as well - Here's the measurements I'm getting compared to the IIR filters I was using. How would you interpret these differences?

View attachment 190611

I guess you did at least a 9-point measurement? According to my experience it is really important to do the measurement carefully and thoroughly since the result will be as precise as the measurement method itself

Would be great to zoom in on around the 0ms part but so far I would say:
- the rise time of the step response is faster with Dirac - you can see that before 0ms. In practice this shall give you better transient response
- X axis zero crossing happens earlier which tells you about the bass characteristics - I am just guessing but either you have less bass or more precise bass by the looks of it
- ringing is somewhat less

What you can also do is to take a look at the IR curve by the function of frequency (Wavelet diagram), you can do that in the Spectrogram by setting it as the below:

1646508879500.png


This graph will tell you about the 'speed' of each frequency (some people here will probably kill me for using this word....)
Anything happens before 0ms is the rise-time and any eventual pre-ringing whereas anything after 0ms is post-ringing - both of them shall be as short (=fast) as possible
You could compare your miniDSP measurement with Dirac using this graph too

I hope this helps
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,110
Likes
2,327
Location
Canada
I've been looking to (and working on) improve the stereo setup in my new place for a while now; Mainly moving the speakers around, applying EQ (on a miniDSP-like setup), and recently (making quite a large impact) moving aside my large oak coffee table. The current setup is quite basic:

View attachment 190319

On the left is open space for ~5 meters, on the right is a hard wall / window with curtains, and behind the couch I'm sitting there's about 1 meter of space before the back wall. There's no room treatment besides a 200 cubic feet foam bean bag against that wall. The current (MMM) frequency response over the 3 seater sofa (with EQ) looks like this:

View attachment 190320

I'm still not quite happy with the sound as it is - I've heard many better systems, but am not really sure on what to improve from this point onward. I've been looking at step responses and impulse responses, but can't quite find information on what these should look like and how (significantly) that influences the sound. The same goes for waterfall graphs and spectrograms.

Room treatment wise I read that rugs are heavily recommended, thick absorption behind the speakers 'helps', ceiling treatment is good, side reflections can kinda stay, and generally filling your walls with panels helps dampening the room - but - I find it hard to link those improvements to actual measurements (and thus keep track of whether it's objectively helping).

Could some of you provide some guidance on these time / frequency domain measurements translate to perceived sound? I'd love to be able to work towards measurable goals in treating the room / perhaps purchasing new speakers / bi-amping these to time align the drivers (if that does anything significant). I've included my REW measurements of the room as it is.

Many thanks!
- David

Appendix pictures:
View attachment 190323

View attachment 190327

View attachment 190325

Polarity looks inverted to me. Also, I've noticed that with some of these automated time corrections... while the wavelet's energy peak looks "correct", the phase response is overshooting well past the minimum phase. Is that supposed to be a good thing?

*On closer inspection, it does seem to be overshooting the zoomed-in spectrograms as well. However, with the long decay I do wonder if this is even audible.

1646525007539.png
 
Last edited:

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,110
Likes
2,327
Location
Canada
You have some weird distortion peaks at 1000, 2000, 3000, and 4000Hz. Is there a problem with the measurement mic, or the amp powering the speakers?

I suspect the visible 1kHz harmonic peaks are mostly from the microphone. I notice them sometimes appearing in my own measurements every now and again for a few hours of the day. Even after turning off all other electrical equipment and speakers, it persistently remains -- until it just suddenly goes away. I haven't figured out where it's picking up the noise from, though.
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,110
Likes
2,327
Location
Canada
@Naughtius

How was the correction made in your original posted measurements exactly? Could you also attach the uncorrected individual left and right measurements along with the the results from the new Dirac corrected IRs? I'm curious to see what the difference between them...
 

Hipper

Addicted to Fun and Learning
Joined
Jun 16, 2019
Messages
753
Likes
625
Location
Herts., England
Here's a pic of my listening room:

017a.JPG


It's a dedicated room so I could do what I like. I like the sound. Mine's an extreme example but I got there over a few years. In particular I did not like side wall reflections hence the panels each side of the speakers. Some people like side wall reflections. I sit upright in a that chair but with my feet up. I also sit in the semi darkness.

My set up is with a single sweetspot. It's a bit more difficult for two or three listening positions but that seems to be why people use the MMM (moving mic measurement) system, or Diracs multi position one.

Going back to your first post, I responded by saying that bass is your first target to correct, the 30-300Hz range. To do this I used solely measurements using frequency response (with 'no smoothing'), waterfalls and spectograms, and finally the Real Traps Test Tone CD of bass frequencies and my ears. You may be able to get this better using only DSP/EQ, or only subs, but when using bass traps you also get the decay times down. I understand DSP/EQ does this too to some extent but I don't know from experience. What I do believe is that lowering decay times is a big factor in the improvements to my sound, especially bass.

The higher frequencies are harder to understand from measurements because they will include not only the direct sound from the speakers but all the various reflections arriving at different times. In this area microphones do not behave like our ears and brain and simply sum these things up. Our brain acts differently (called psychoacoustics - see for example the precedence or Haas effect, equal loudness or Fletcher-Munson curves). My approach was to remove reflections, using absorbent panels, and then look at the frequency response and making broad EQ adjustments to this, like tone controls. The aim was approximately flat but in truth I used mostly my ears to get it to sound good to me. I struggled at first to get good detail in the percussion but eventually managed it.

Looking at your photo in your first post, the position of the speakers and the way you listen (judging by your leg position) won't help with what you hear.

Symmetry is an important requirement, each speaker ideally being equally distant from the side and back walls and better still in a symmetrical room. Your right one is very close to the corner, your left is not. This will affect bass and higher frequencies. Indeed if you imagine a ray of light coming from your right speaker (which is how the frequencies over about 300Hz typically behave), reflecting off the side wall like from a mirror and then to your ears and compare that with how a similar ray would arrive at your ear from the left speaker you might guess, rightly, that the left one has longer to travel and would arrive later. That would mess up the sound.

Placing speakers in their ideal position may be domestically difficult but you might consider placing them for listening then storing them in the corner when not in use. Same applies for your listening position. An alternative would be to put some absorbent material right beside the right speaker and assume that the left speakers sound doesn't reflect at all. If that doesn't help then an arrangement like mine with panels by both speakers should do it.

If you noticed the difference a coffee table makes then putting a rug on the floor would be a good idea (some say we don't notice floor reflections because we are used to them). Definitely worth experimenting.

On ceiling reflections I'm not convinced. My speakers have a narrow vertical dispersion pattern and don't have any ceiling reflections that I can hear. I did experiment by jamming some insulation (you could try a duvet) on the ceiling with a 'T' shaped piece of wood. Perhaps ask your speaker makers if it might be an issue. Of course how close you sit to the speakers has an impact on this.

'The Thirds' positioning that I gave in my linked post is designed to reduce the impact of bass issues but also reflections. Again worth an experiment even if it's not practical for you long term.

To sum up, symmetry of room and positioning and removing reflections (as well as controlling the bass) means I hear practically only the sound direct from the speakers. My brain then sums this up to create a phantom image somewhere between the two speakers. After that it's the recording that dictates how the sound is spread. Usually I get rock solid central vocals, instruments spread around in the space, sometimes with a bit of depth. I also hear a lot of detail, such as percussion or subtle harmony singing. I do not get imaging wider then the speakers. A sort of good headphone experience but in the space of the room.
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,110
Likes
2,327
Location
Canada
Not an expert, but this lower area around 350-800hz can't be great right?

View attachment 190337
Note: Not any particular target curve, just a quick and dirty visual linear best fit to show that area more easily. Blue line at same slope offset down to easily show about how many dB down on SPL scale.

Assymetry in the left and right response could be better as well as the uneven decay... but, I think that very broad dip in the all-important-mids section is a far larger contributor to the OP's dissatisfaction than anything else.

Is this primarily a fault of the speakers or the room? Perhaps both?

There isn't much of a huge difference in your MMM measurements compared to mine at approximately 2.3 and 1.5 meters:
1646589768601.png


But the mids looks quite deficient -- significantly so in your single point measurements vector averaged above. IMO, you need to investigate what is causing this first...
 
OP
N

Naughtius

Member
Joined
Mar 4, 2022
Messages
56
Likes
70
I guess you did at least a 9-point measurement? According to my experience it is really important to do the measurement carefully and thoroughly since the result will be as precise as the measurement method itself

Would be great to zoom in on around the 0ms part but so far I would say:
- the rise time of the step response is faster with Dirac - you can see that before 0ms. In practice this shall give you better transient response
- X axis zero crossing happens earlier which tells you about the bass characteristics - I am just guessing but either you have less bass or more precise bass by the looks of it
- ringing is somewhat less

What you can also do is to take a look at the IR curve by the function of frequency (Wavelet diagram), you can do that in the Spectrogram by setting it as the below:

View attachment 190642

This graph will tell you about the 'speed' of each frequency (some people here will probably kill me for using this word....)
Anything happens before 0ms is the rise-time and any eventual pre-ringing whereas anything after 0ms is post-ringing - both of them shall be as short (=fast) as possible
You could compare your miniDSP measurement with Dirac using this graph too

I hope this helps

I went for the full 17-point set over the entire sofa - and might indeed have had a little less bass with the Dirac calibration. I've redone the both calibrations now the furniture has moved around a bit and come to the following wavelets:

IIR.jpg
Dirac.jpg


If I'm interpreting these correctly my peak low end energy only arrives after 25ms? These would be reflections from the room coming back and amplifying the signal at the listener position?
 
OP
N

Naughtius

Member
Joined
Mar 4, 2022
Messages
56
Likes
70
@Naughtius

How was the correction made in your original posted measurements exactly? Could you also attach the uncorrected individual left and right measurements along with the the results from the new Dirac corrected IRs? I'm curious to see what the difference between them...
There are IIR filters from REW loaded into an ADAU1701 chip - same chip as in the smallest MiniDSP. Everything has moved around a bit, so I've redone the measurements with a new Dirac calibration and attached the files. :)
 

Attachments

  • Karl Analysis Measurements R Raw-Filtered-Dirac.zip
    3.5 MB · Views: 74
  • Karl Analysis Measurements L Raw-Filtered-Dirac.zip
    3.5 MB · Views: 71

EEE272

Active Member
Joined
Jan 4, 2022
Messages
128
Likes
61
I went for the full 17-point set over the entire sofa - and might indeed have had a little less bass with the Dirac calibration. I've redone the both calibrations now the furniture has moved around a bit and come to the following wavelets:

View attachment 190826View attachment 190827

If I'm interpreting these correctly my peak low end energy only arrives after 25ms? These would be reflections from the room coming back and amplifying the signal at the listener position?
It is perfectly normal that the lower frequencies take a bit longer. 25 ms should not be perceivable.

I am still thinking that your imaging will benefit from the speakers being farther away from the wall. I think you currently might have some nearby reflection on the adjacent wall that is canceling out your direct sound (SBIR). If you move the speakers, you could check if the dip moves as well. This would be an artifact that you cannot correct via EQ nor in the time domain. In the example room that you tested, you see that the speakers also have some distance from the back wall and there are absorbers to help with the effect even further. "Messy" might also sound like interference.
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,110
Likes
2,327
Location
Canada
There are IIR filters from REW loaded into an ADAU1701 chip - same chip as in the smallest MiniDSP. Everything has moved around a bit, so I've redone the measurements with a new Dirac calibration and attached the files. :)

BTW, you have accidentally mislabeled (or switched) the Dirac LR measurements.

Corrections by Dirac looks good. Although, the HF level looks a tad bit off -- could be caused by the microphone position.

There isn't much of a noticeable difference in the wavelets except improvement in frequency magnitude evenness.

test wave left.gif test wave right.gif

Everything has moved around a bit

Hmmn... So the cause of broad dip in the mids appears to be bad positioning, then. Glad that mystery was solved! :)

Now, delayed energy esp. between 200-600 Hz is from boundary reflections which can be easily treated with acoustic absorption -- which Hipper has kindly illustrated above. *You probably do not really need that much acoustic absorption, of course. :p


FDW 5 cycles
1646601938040.gif
 
Last edited:
OP
N

Naughtius

Member
Joined
Mar 4, 2022
Messages
56
Likes
70
Here's a pic of my listening room:

View attachment 190766

It's a dedicated room so I could do what I like. I like the sound. Mine's an extreme example but I got there over a few years. In particular I did not like side wall reflections hence the panels each side of the speakers. Some people like side wall reflections. I sit upright in a that chair but with my feet up. I also sit in the semi darkness.

My set up is with a single sweetspot. It's a bit more difficult for two or three listening positions but that seems to be why people use the MMM (moving mic measurement) system, or Diracs multi position one.

Going back to your first post, I responded by saying that bass is your first target to correct, the 30-300Hz range. To do this I used solely measurements using frequency response (with 'no smoothing'), waterfalls and spectograms, and finally the Real Traps Test Tone CD of bass frequencies and my ears. You may be able to get this better using only DSP/EQ, or only subs, but when using bass traps you also get the decay times down. I understand DSP/EQ does this too to some extent but I don't know from experience. What I do believe is that lowering decay times is a big factor in the improvements to my sound, especially bass.

The higher frequencies are harder to understand from measurements because they will include not only the direct sound from the speakers but all the various reflections arriving at different times. In this area microphones do not behave like our ears and brain and simply sum these things up. Our brain acts differently (called psychoacoustics - see for example the precedence or Haas effect, equal loudness or Fletcher-Munson curves). My approach was to remove reflections, using absorbent panels, and then look at the frequency response and making broad EQ adjustments to this, like tone controls. The aim was approximately flat but in truth I used mostly my ears to get it to sound good to me. I struggled at first to get good detail in the percussion but eventually managed it.

Looking at your photo in your first post, the position of the speakers and the way you listen (judging by your leg position) won't help with what you hear.

Symmetry is an important requirement, each speaker ideally being equally distant from the side and back walls and better still in a symmetrical room. Your right one is very close to the corner, your left is not. This will affect bass and higher frequencies. Indeed if you imagine a ray of light coming from your right speaker (which is how the frequencies over about 300Hz typically behave), reflecting off the side wall like from a mirror and then to your ears and compare that with how a similar ray would arrive at your ear from the left speaker you might guess, rightly, that the left one has longer to travel and would arrive later. That would mess up the sound.

Placing speakers in their ideal position may be domestically difficult but you might consider placing them for listening then storing them in the corner when not in use. Same applies for your listening position. An alternative would be to put some absorbent material right beside the right speaker and assume that the left speakers sound doesn't reflect at all. If that doesn't help then an arrangement like mine with panels by both speakers should do it.

If you noticed the difference a coffee table makes then putting a rug on the floor would be a good idea (some say we don't notice floor reflections because we are used to them). Definitely worth experimenting.

On ceiling reflections I'm not convinced. My speakers have a narrow vertical dispersion pattern and don't have any ceiling reflections that I can hear. I did experiment by jamming some insulation (you could try a duvet) on the ceiling with a 'T' shaped piece of wood. Perhaps ask your speaker makers if it might be an issue. Of course how close you sit to the speakers has an impact on this.

'The Thirds' positioning that I gave in my linked post is designed to reduce the impact of bass issues but also reflections. Again worth an experiment even if it's not practical for you long term.

To sum up, symmetry of room and positioning and removing reflections (as well as controlling the bass) means I hear practically only the sound direct from the speakers. My brain then sums this up to create a phantom image somewhere between the two speakers. After that it's the recording that dictates how the sound is spread. Usually I get rock solid central vocals, instruments spread around in the space, sometimes with a bit of depth. I also hear a lot of detail, such as percussion or subtle harmony singing. I do not get imaging wider then the speakers. A sort of good headphone experience but in the space of the room.
Thanks a lot for this elaborate post (as well as for your others I've read) - really helps in getting a feel what matters and what to work on - and awesome to see your setup! I've quickly been experimenting with various forms of absorption and feel like I'm getting to the root cause of my discontent - I'll write more down below.

BTW, you have accidentally mislabeled (or switched) the Dirac LR measurements.

Corrections by Dirac looks good. Although, the HF level looks a tad bit off -- could be caused by the microphone position.

There isn't much of a noticeable difference in the wavelets except improvement in frequency magnitude evenness.

View attachment 190850 View attachment 190851



Hmmn... So the cause of broad dip in the mids appears to be bad positioning, then. Glad that mystery was solved! :)

Now, delayed energy esp. between 200-600 Hz is from boundary reflections which can be easily treated with acoustic absorption -- which Hipper has kindly illustrated above. *You probably do not really need that much acoustic absorption, of course. :p


FDW 5 cycles
View attachment 190855
Awesome GIFs, thanks for the help and diving in! You're getting me excited about buying a few packs of rockwool / scraps of wood / roll off fabric and starting to experiment with actually treating the room.
 
OP
N

Naughtius

Member
Joined
Mar 4, 2022
Messages
56
Likes
70
I've also already been diving into the reflections from the impulse response envelope from my first post, and it looks like I've found the main cause of my setup sounding 'messy' despite the decent frequency response.

Here's most of the reflections from the first setup identified and mitigated:
Reflections 2.jpg


For this I have:
- Placed pillows on the floor between the LP and the speakers
- Closed the curtains at the front window
- Put a mattress horizontally against the back wall at listening height
- Closed the right wall curtains
- Placed a pillow 50cm above the mic to block the ceiling reflections

I've attached an MP3 with a recording of a random drum sequencer, playing in both situations. Please don't mind the background noise. Curious if you hear the difference as clearly as I do here!
 

Attachments

  • Reflection management differences.zip
    392.1 KB · Views: 90
Last edited:

TomJ

Active Member
Joined
Sep 16, 2019
Messages
129
Likes
178
Location
Palo Alto CA
Do you have any room treatment / have you ever looked at the time domain in your room?
David, nice analysis and mitigation.
FYI here's how DRC alone with no room treatment or furniture changes improved the temporal response of the same sound system 2 years ago in our previous place, a "concrete box" with lots more acoustic problems (eg top chart) than our current place. I just found those .mdat files and generated ETC charts on the IR before and after DRC (middle, bottom). I never looked at the IR before, surprised actually by how much Dirac improved it in such a difficult space.
cheers, Tom

FR and IR ETC, DRC on vs off.png
 
Last edited:
OP
N

Naughtius

Member
Joined
Mar 4, 2022
Messages
56
Likes
70
Clear difference even playing the audio on my phone.
It's night and day for me as well! I find it curious no one pointed this out from the measurements, but the general consensus seems to be frequency response first. No one would accept a (time) distortion like this from their speakers or their amplifier, but in a room it's apparently not common to look at these reflections first
 
OP
N

Naughtius

Member
Joined
Mar 4, 2022
Messages
56
Likes
70
David, nice analysis and mitigation.
FYI here's how DRC alone with no room treatment or furniture changes improved the temporal response of the same sound system 2 years ago in our previous place, a "concrete box" with lots more acoustic problems (eg top chart) than our current place. I just found those .mdat files and generated ETC charts on the IR before and after DRC (middle, bottom). I never looked at the IR before, surprised actually by how much Dirac improved it in such a difficult space.
cheers, Tom

View attachment 191105
Hey Tom! Sweet looking improvement, especially the frequency response.

For the ETC I think we're not actually looking at less reflections, only a more sharply defined first peak, and thus when scaled to % it looks like the rest is all attenuated - but I could be wrong here.

Could you attach the REW files? Curious to see what the raw measurements look like
 
Top Bottom