2.1 Stereo Calibration - what am I missing?

[Dave the Rave]

I did further measurements without the Acoustic Timing as attached. One positive is I got the Subwoofer polarity corrected (had it wrong all this time where I was losing some mid-bass). Since the right speaker distortion only happens at high volume, I turned down the mid and high of both speakers to 10 o'clock where the distortion is faintly noticeable.
As usual I measured with and without subs.
I will do some listening using the applied EQ on R & L measurements (without Sub this time). .Let me know if the EQ looks OK to you.
I want to keep the speaker repair as last resort..

 

[John Mulcahy]

Measurements look better now, none of them show any distortion issues though there is some 2nd harmonic around 800 Hz on the R channel. All the measurements with the sub have a lot of pre-ringing, especially the first. I wonder whether the measurement issues had more to do with the convolver than the system.

 

[Dave the Rave]

Thanks John Any advice how to identify the cause or address the pre-ringing issue (not much room to move sub though)?

 

[John Mulcahy]

I imagine it must be the filters that are being applied to it.

 

[Dave the Rave]

After much reading and help in here, I think I managed to properly measure the speaker and level matching them (playing with the squawker and balance level). Also put some damping behind the sub.
I applied the EQ (L2 and R2 measures as attached) - have to say this is the best sounding configuration I have so far. Imaging is excellent and sound is crisp.
Is there anything I can do for the 50 and 500 hz dips (maybe to get more sustained bass)?

 

[JStewart]

I think I managed to properly measure the speaker

The ringing is gone from the impulse responses. Any idea what what the issue was with the prior measurements?

have to say this is the best sounding configuration I have so far. Imaging is excellent and sound is crisp.

Come a long way from the frequency response graph in the OP.

 

[Dave the Rave]

I suppose bringing the speaker to level match made a huge difference. Not sure how the ringing was gone and how too identify it.
The last bit I want to improve is the low bass area for instruments such as the snare..... at times I hear too much bass. Not sure what can be done and if the big dips at 50 and 300 hz are issues I need to fix.

 

[JStewart]

The last bit I want to improve is the low bass area for instruments such as the snare..... at times I hear too much bass.

I can’t say what you’re hearing, but you may want to consider there are zero standards for recording levels and gear used. So some material is going to sound good and some not. Google Floyd Toole Circle of Confusion for detailed explanation.

Not sure what can be done and if the big dips at 50 and 300 hz are issues I need to fix.

I’d focus on the 50 1st. It’s a wide dip in that region.

 

[ddude003]

Not sure how the ringing was gone and how too identify it.

What Is Ringing in Audio (Artifacts)? [Definitive Guide]

Discover the secrets of audio ringing and its impact on sound quality. What causes the distortion? How to minimize it, linear, minimum phase, apodizing filters?

samplerateconverter.com

You might try and gain match your subs to your mains to tone down that bass a bit... And as @JStewart suggests above, the quality of recording, mixing and mastering varies wildly... An example database: https://dr.loudness-war.info

 

[moedra]

I downloaded the DAC4.mdat file. This is with the controls on the two speakers set to where they will stay, yes? Seems like the speakers measure well now. If you want to experiment a little, and if you'd like to attach your house curve, I can make a filter for you, and we can see what happens.

 

[Dave the Rave]

I downloaded the DAC4.mdat file. This is with the controls on the two speakers set to where they will stay, yes? Seems like the speakers measure well now. If you want to experiment a little, and if you'd like to attach your house curve, I can make a filter for you, and we can see what happens.

Thank you. Yes its for the 2 speakers- R1 and R2 means I ran the measure twice for the right speaker... (hope I understood the question correctly).
Attached is the House Curve I am using.
I will be uploading the filters using a WAV convolver through Daphile's BruteFIR with a Left and ight channel file

 

[moedra]

Thank you. Yes its for the 2 speakers- R1 and R2 means I ran the measure twice for the right speaker... (hope I understood correctly the question).
Attached is the House Curve I am using.
I will be uploading the filters using a WAV convolver through Daphile's BruteFIR as a stereo Left/Right channel file

This was made from L1 and R1 with a linear target. I'll make another one using your curve.

Here is the simulated frequency response of this filter:

And this is the simulated step response:

You should get something close to this, at least.

 

[moedra]

Thank you. Yes its for the 2 speakers- R1 and R2 means I ran the measure twice for the right speaker... (hope I understood the question correctly).
Attached is the House Curve I am using.
I will be uploading the filters using a WAV convolver through Daphile's BruteFIR with a Left and ight channel file

Here's a filter made using the house curve as the target. As before I've just included images of the simulated frequency and step responses for reference. The frequency response includes the original measured L/R responses in grays. Let me know how these filters sound.

 

[Dave the Rave]

Thanks heaps Moedra. I will do some critical listening and report back. You are too kind.

 

[Dave the Rave]

Hi Moedra, The filter made a really nice difference in my system. Felt like it opened a door where some of the hidden instruments can now be clearly heard. Voices are much clearer with extended top end. Bass especially is tighter.
Only observations - The volume need to be turned up higher (not a biggie though) and some songs had clipping notice from Daphile. Last element is the bass which I find much more controlled now but would not mind a bit more of it.
My missus is getting the blind behind the speakers changed this week so I would need to re-measure again. If you let me know how to apply the filter, I can give it a try.
All in all, I really like that filter so thanks again.

 

[moedra]

Hi Moedra, The filter made a really nice difference in my system. Felt like it opened a door where some of the hidden instruments can now be clearly heard. Voices are much clearer with extended top end. Bass especially is tighter.
Only observations - The volume need to be turned up higher (not a biggie though) and some songs had clipping notice from Daphile. Last element is the bass which I find much more controlled now but would not mind a bit more of it.
My missus is getting the blind behind the speakers changed this week so I would need to re-measure again. If you let me know how to apply the filter, I can give it a try.
All in all, I really like that filter so thanks again.

Thanks! Believe me when I tell you that I appreciate positive feedback. It means I haven't been wasting my time. Creating these filters has become sort of a thing of passion for me. The process I use within REW is one I have been working on for the past year and a half or so. It has been a very time-consuming process of trial and error, research and more trial and error. Over-complicating things and simplifying things. Comparing results to professional packages and intense listening sessions. Starting over and over again, reaching out to John and Mitch for feedback, and keeping track of changes. Trying different approaches to measuring and breaking things down. What you're listening to is the result of all that.

Just a few thoughts:

• A reduction in level is inherent as a result of the attenuation of peaks. This is necessary because the nulls cannot be boosted. That is par for the course as they say. You should need to raise the level at the amplifier by 6 to 8 dB to compensate. One of the things I have no control over is how your convolver plays back the level. Does it normalize? If so, deactivate level normalization. You may need to manually reduce the level 1 or 2 dB at the source if possible to prevent that clipping notification.

• Which filter are you referring to here? The linear one I sent you first or the second one I made with your house curve? I'm guessing you're talking about the linear one. Have you compared both filters? If you can give me pros and cons with each one I can tailor a third filter to your preference once you measure again.

• How would you describe the soundstage? Does it sound bigger/wider or smaller/narrower than it was before? Does it sound the same as before?

• The procedure is simple enough in concept but a bit lengthy. Half of it is pretty technical and the other half is more of an art form that you have to train yourself for. It's one of those things that can be learned in an hour but takes much longer to master. If you're interested I can give you some insight and pointers to help you get started.

• When you do measure again, make sure that the sweep plays at a reasonably loud volume, but not so loud that it hurts your ears or damages your speakers. A good level is around 75 to 80 dB (SPL). Ensure that the left and right sweeps are clean, with no driver distortion or resonance noise. You do not need to sweep both speakers together. Do them one at a time.

• Point your mic up at the ceiling when measuring, If you have a 90° calibration file, use it!

 

[sm52]

Do you have words to describe the changes you've became to your speakers using REW's features? What has become less audible? What has become audible that you have not heard before? I applaud the work you have done.

 

[moedra]

Do you have words to describe the changes you've became to your speakers using REW's features? What has become less audible? What has become audible that you have not heard before? I applaud the work you have done.

Are you asking me or the thread starter?
If you're asking me, the changes made to the frequency response balance the channels and solidify the phantom center, which is evident in the tightened bass and transient response Dave is describing.

What has become less audible?
Distortions caused by acoustic anomalies in the frequency response.

What has become audible that has not been heard before?
Details that have not been smeared by timing and frequency imbalances.

 

[ddude003]

@moedra, Are you using Audiolense or only REW to produce your filters? Have you tried or compared Serkan Gur's inversion methods, as was previously pointed out above?

 

[moedra]

@moedra, Are you using Audiolense or only REW to produce your filters? Have you tried or compared Serkan Gur's inversion methods, as was previously pointed out above?

Just REW and rePhase. The goal has always been to replicate what I hear in filters generated from other pro packages. I have been focused on what I hear in the results.

Yes I have been through Serkan's videos and I have been through his and others' procedures. I recall there being something about each one of them that I didn't like in the results, which helped me to narrow down my own procedure. I do have Audiolense, as well as Acourate, Focus Fidelity, ARC3, and Dirac Live. I studied them to help me with REW also. I have been doing a great deal of listening and A/B comparisons since February of last year. The biggest breakthroughs for me were discovering how to best align the step response and realizing I had access to multiple EQ banks in rePhase. I was having difficulty keeping the stereo image open and wide for the longest time, but I finally worked that out. The last hurdle was the bass response. The state of the art packages all provide a deep and detailed bass response, and until recently I hadn't figured out how to reach that point.

If you'd like a demo filter, post a .mdat with a single calibrated measurement from each channel taken from your mlp. Doing that will help me, too, because it lets me gather information about how the filters perform in different systems. Keep in mind that I am currently only doing 2-ch stereo... I'm working on adding sub filters for multiple outputs but I'm not there yet.

If you're on Windows, you can actually use Focus Fidelity's Impala to take your measurements. The new version David released a week or so ago supports clock drift compensation, so if you get the pulses aligned (mic dead center) when measuring, it makes creating filters in REW very efficient. Impala is also much faster than REW when measuring, and can export its measurements as WAV files, which REW can import.

 

[Dave the Rave]

I do realise the task must be really arduous to achieve what you created - I am a newbie on Rew/technical measurements and have already spent hours and hours but came nowhere close to those filters - so appreciate the hard work you have put in over years and sharing your passion. In regards to the questions:

• A reduction in level is inherent as a result of the attenuation of peaks. This is necessary because the nulls cannot be boosted. That is par for the course as they say. You should need to raise the level at the amplifier by 6 to 8 dB to compensate. One of the things I have no control over is how your convolver plays back the level. Does it normalize? If so, deactivate level normalization. You may need to manually reduce the level 1 or 2 dB at the source if possible to prevent that clipping notification.
- Understand the peaks with the corrections. I am assuming BruteFir convolver does not normalise as the "Match Filter Gain" is checked off. I will try this setting on next time i get the clipping (only happens for few songs esp on Spotty). Daphile does not seem to have a feature to control the level at source .

• Which filter are you referring to here? The linear one I sent you first or the second one I made with your house curve? I'm guessing you're talking about the linear one. Have you compared both filters? If you can give me pros and cons with each one I can tailor a third filter to your preference once you measure again.
I tried the linear filter temporarily and switched to the house curve filter for most listening as i got carried away with the sudden nice change of either filters. Today I did more critical listening through the first linear filter and definitely prefer this one. Below is my impression of both filters:
- Overall Clarity - House Curve seems blacker by a very small margin maybe due to louder low end. But linear has a bit more granular details.
- Soundstage - Linear gives wider soundstage with all instruments noticeable. Voices are well centered (with instruments layered well in the centre/Right/Left and how far from the singer. House curve has the same feeling but again with the higher volume of linear, the latter gave the impression of having an edge
- Low/High - Low freq firmer for linear. House Curve louder low freq and high freq seems slightly better rolled off.
- Listening experience - both filters are equally excellent with no harshness or loudness or being too flat and boring. Could listen for hours either at low or high volume
-Clipping - Hardly any clipping with Linear . Slightly more with House Curve with all types of files (more with Spotty)

• How would you describe the soundstage? Does it sound bigger/wider or smaller/narrower than it was before? Does it sound the same as before?
- Feels the soundstage is less wide than before but felt some instruments were played too wide previously.... the filters brought them together,.

• The procedure is simple enough in concept but a bit lengthy. Half of it is pretty technical and the other half is more of an art form that you have to train yourself for. It's one of those things that can be learned in an hour but takes much longer to master. If you're interested I can give you some insight and pointers to help you get started.
- thanks for the offer. Given the result, I am not sure i want to tinker too much now. I am sure I will create other issues if i start changing stuffs again..

• When you do measure again, make sure that the sweep plays at a reasonably loud volume, but not so loud that it hurts your ears or damages your speakers. A good level is around 75 to 80 dB (SPL). Ensure that the left and right sweeps are clean, with no driver distortion or resonance noise. You do not need to sweep both speakers together. Do them one at a time.
Noted. My amp and speakers play really loud hence 75db sound too loud (at 1/4 volume) hence I stick to 70db. In the next round, should I measure the Left/Right with and without the sub and have a separate measurement for the sub?
Previously I measured with the sub always on as the sub is fed through the pre-out of the amp without any active crossover control feature.


• Point your mic up at the ceiling when measuring, If you have a 90° calibration file, use it!
Yes I always use the UMIK-1 90 degree file with the mic as described.

 

[Dave the Rave]

Do you have words to describe the changes you've became to your speakers using REW's features? What has become less audible? What has become audible that you have not heard before? I applaud the work you have done.

Please refer to my summary above to describe the changes. Lets say I am hearing more details that always seem to be there but now with better separation and transient . I listened to some tracks which were hard to listen before (e.g Lover in Japan by Coldplay or Extreme Ways by Moby) and now they sound way better.

 

[moedra]

I do realise the task must be really arduous to achieve what you created - I am a newbie on Rew/technical measurements and have already spent hours and hours but came nowhere close to those filters - so appreciate the hard work you have put in over years and sharing your passion. In regards to the questions:

• A reduction in level is inherent as a result of the attenuation of peaks. This is necessary because the nulls cannot be boosted. That is par for the course as they say. You should need to raise the level at the amplifier by 6 to 8 dB to compensate. One of the things I have no control over is how your convolver plays back the level. Does it normalize? If so, deactivate level normalization. You may need to manually reduce the level 1 or 2 dB at the source if possible to prevent that clipping notification.
- Understand the peaks with the corrections. I am assuming BruteFir convolver does not normalise as the "Match Filter Gain" is checked off. I will try this setting on next time i get the clipping (only happens for few songs esp on Spotty). Daphile does not seem to have a feature to control the level at source .

• Which filter are you referring to here? The linear one I sent you first or the second one I made with your house curve? I'm guessing you're talking about the linear one. Have you compared both filters? If you can give me pros and cons with each one I can tailor a third filter to your preference once you measure again.
I tried the linear filter temporarily and switched to the house curve filter for most listening as i got carried away with the sudden nice change of either filters. Today I did more critical listening through the first linear filter and definitely prefer this one. Below is my impression of both filters:
- Overall Clarity - House Curve seems blacker by a very small margin. House Curve plays lower than linear hence maybe the reason.
- Soundstage - Linear gives wider soundstage with all instruments noticeable. Voices are well centered (with instruments layered well in the centre/Right/Left and how far from the singer. House curve has the same feeling but again with the higher volume of linear, the latter gave the impression of having an edge
- Low/High - Low freq firmer for linear. House Curve louder low freq and high freq seems better rolled off. Would love something in between.
- Details - prefer the Linear maybe because volume seems higher than House Curve
- Listening experience - both filters are equally excellent with no harshness or loudness or being too flat and boring. Could listen for hours either at low or high volume
-Clipping - Extremely few samples with Linear mainly with Spotty. Slightly more with House Curve with all types of files

• How would you describe the soundstage? Does it sound bigger/wider or smaller/narrower than it was before? Does it sound the same as before?
- Feels the soundstage is less wide than before but felt some instruments were played too wide previously.... the filters brought them together,.

• The procedure is simple enough in concept but a bit lengthy. Half of it is pretty technical and the other half is more of an art form that you have to train yourself for. It's one of those things that can be learned in an hour but takes much longer to master. If you're interested I can give you some insight and pointers to help you get started.
- thanks for the offer. Given the result, I am not sure i want to tinker too much now. I am sure I will create other issues if i start changing stuffs again..

• When you do measure again, make sure that the sweep plays at a reasonably loud volume, but not so loud that it hurts your ears or damages your speakers. A good level is around 75 to 80 dB (SPL). Ensure that the left and right sweeps are clean, with no driver distortion or resonance noise. You do not need to sweep both speakers together. Do them one at a time.
Noted. My amp and speakers play really loud hence 75db sound too loud (at 1/4 volume) hence I stick to 70db. In the next round, should I measure the Left/Right with and without the sub and have a separate measurement for the sub?
Previously I measured with the sub always on as the sub is fed through the pre-out of the amp without any active crossover control feature.


• Point your mic up at the ceiling when measuring, If you have a 90° calibration file, use it!
Yes I always use the UMIK-1 90 degree file with the mic as described.

I agree that the linear filter is preferable. The slope is +2.3dB at 10Hz and -2.3dB at 24kHz. In all my tests I simply have not been as satisfied by anything other than that target. To my ears it is perfectly balanced. As soon as I start bumping things around, trying to compensate for what is described as perceived linearity, I start noticing that certain frequencies get masked and hidden. The bass boost in that Harmon curve, for instance. It caused the bass to feel overly heavy. That's a 6dB lift in the bottom there! It is only when the target is flat and the tilt is just right that everything gels and opens up.

Level-matching is important! We perceive even the most subtle increase in perceived volume as an improvement, so if possible be sure to listen to each filter at the same playback level. When switching between them you should only hear tonal and other character-oriented changes. If the volume seems to jump, they aren't matched yet.

With regard to the soundstage, it is possible to have a presentation that is too wide. That's a phase issue. If the soundstage is too wide, it's likely that there seemed to be a hole in the center, especially if too much of the signal was out of phase. The fact that it is more accurate and put together now, even though it sounds somewhat narrower than before, should be a good thing.

Next round, measure the same way. Leave the sub on just the way it is. You can't have a separate filter for the sub unless you have independent control of the sub on its own channel. Have you got an extra pair of outputs (3/4 for example) that you could use to drive the sub?

 

[Dave the Rave]

I am listening to both filters right now interchangeably through another player (LMS on NAS playing on an Oppo 103 which then goes to my DAC) and now using InguzEQ (instead of BruteFIR). I now find the volume the same for both the House Curve and linear. The difference is EXACTLY what you described. More pronounced bass for house curve which I don't like anymore as I got used to the more controlled low end and details of linear. You are right, you cant mess around with the linear as you lose something else. For the soundstage, since all come together now , speakers seem to have disappeared (i.e in the past I would see an emphasis of an instrument on the left or right speaker which could be distracting at times). I prefer the whole enveloped sound as the linear filter provides.
I am not sure of the question on sub - I do not have another amp to drive the sub. My DAC has another unbalanced output but i think its fixed 100% volume so not workable. Will get the measurements done once the blinds are installed.

FYI, I am so pleasantly surprised I played around with the different files. My Oppo only output 96kpbs hence was using the 96k convolver. But when I used the `192k file, I had another jump on the resolution which i absolutely love (though the 192k file seems to be downscaled back to 96k as shown on my DAC). I am not sure I should try higher than 192k though as my DAC goes to that limit .

 

[moedra]

I am listening to both filters right now interchangeably through another player (LMS on NAS playing on an Oppo 103 which then goes to my DAC) and now using InguzEQ (instead of BruteFIR). I now find the volume the same for both the House Curve and linear. The difference is EXACTLY what you described. More pronounced bass for house curve which I don't like anymore as I got used to the more controlled low end and details of linear. You are right, you cant mess around with the linear as you lose something else. For the soundstage, since all come together now , speakers seem to have disappeared (i.e in the past I would see an emphasis of an instrument on the left or right speaker which could be distracting at times). I prefer the whole enveloped sound as the linear filter provides.
I am not sure of the question on sub - I do not have another amp to drive the sub. My DAC has another unbalanced output but i think its fixed 100% volume so not workable. Will get the measurements done once the blinds are installed.

FYI, I am so pleasantly surprised I played around with the different files. My Oppo only output 96kpbs hence was using the 96k convolver. But when I used the `192k file, I had another jump on the resolution which i absolutely love (though the 192k file seems to be downscaled back to 96k as shown on my DAC). I am not sure I should try higher than 192k though as my DAC goes to that limit .

You're going to want to use the filter resolution that matches your playback sample rate. This way, you won't have any real-time resampling happening.