-
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!
New set of graphs for all speakers reviews
- Thread starter pierre
- Start date
20 bands of filters I assume limits it to software like Equalizer APO. The woes of being a Mac user.
MiniDSP is 10 bands for instance. I don’t own DIRAC but can you do filter bands and 20 at that?
_____
I see some pretty sharp filters, I wonder how that would take to some speakers. I have read many times that sharp filters in the treble could cause issues (including very precise filters, like on the Infinity RC263), but I don’t know how true that is. I do know that if distortion is already pretty high, you may not want to apply large boosts.
Since I believe @amirm does distortion measurements relatively quickly, it would be interesting to see the distortion after he applies corrective EQ in Roon.
A practical issue that unfortunately is out of anyone’s hands is production consistency, the Vanatoo for instance has been shown to have wildly difference responses in the treble, Amir’s sample had a large rise in the treble whereas other models aren’t as extreme. Still, this should be a good starting ground for folks, an impressive undertaking to say the least.
Last edited:
Balle Clorin
Major Contributor
- Joined
- Dec 26, 2017
- Messages
- 1,330
- Likes
- 1,209
Fatastic work!!! Very very greateful, thank you
Some info about my methodology.
All the EQs use 20 filters for accuracy's and consistency's sake.
The filters only affect frequencies from 300 Hz and up. This is to avoid messing with the low frequencies, which the room dominates, and therefore needs individual equalization depending on the dimensions of the room the speakers are being used in.
In order to avoid clipping, the preamp should be lowered by an amount equivalent to the highest boost. I'll see if I can come up with some recommendations soon.
I initially based my EQs around the LW, which I would equalize to flat, but after discovering that certain speakers had their post-EQ predicted preference score lowered, partially due to a reduced SM_PIR, I switched to a PIR based approach, which improves all score components (except for LFX on speakers with an overall lack of energy in the midrange and/or treble) for all speakers. After a bit of experimentation, I settled on a -1.0 dB/octave slope for the PIR EQs. The ideal slope depends on the directivity of the speaker in question, but a fall slope of 1 dB proved to be a fast and efficient one-size-fits-all target, as can be seen by improved NBD_ON, NBD_PIR, and SM_PIR, as well as improved aggregated scores, for every single one of the 60+ models that it was applied to: https://docs.google.com/spreadsheet...IYFRpkY45KxBFTZxYuK-lXXeAz7n3XQYojlaD/pubhtml
The model is meant to be used to predict the preference scores of different speakers, though, not a single speaker with and without EQ. There's a lengthy and on-going discussion about the optimal approach to EQing: direct sound vs PIR. I've listened to the arguments from both sides and decided I'll go back to a LW approach, this time with a -0.2 dB/oct. slope. If a resulting EQ profile does not improve all score components (excluding LFX), a PIR based replacement will be made.
Regardless of your stance in the debate, I can say with a high degree of confidence that my EQs improve the fidelity of your speakers, assuming low tolerances from the manufacturers.
All the EQs use 20 filters for accuracy's and consistency's sake.
The filters only affect frequencies from 300 Hz and up. This is to avoid messing with the low frequencies, which the room dominates, and therefore needs individual equalization depending on the dimensions of the room the speakers are being used in.
In order to avoid clipping, the preamp should be lowered by an amount equivalent to the highest boost. I'll see if I can come up with some recommendations soon.
I initially based my EQs around the LW, which I would equalize to flat, but after discovering that certain speakers had their post-EQ predicted preference score lowered, partially due to a reduced SM_PIR, I switched to a PIR based approach, which improves all score components (except for LFX on speakers with an overall lack of energy in the midrange and/or treble) for all speakers. After a bit of experimentation, I settled on a -1.0 dB/octave slope for the PIR EQs. The ideal slope depends on the directivity of the speaker in question, but a fall slope of 1 dB proved to be a fast and efficient one-size-fits-all target, as can be seen by improved NBD_ON, NBD_PIR, and SM_PIR, as well as improved aggregated scores, for every single one of the 60+ models that it was applied to: https://docs.google.com/spreadsheet...IYFRpkY45KxBFTZxYuK-lXXeAz7n3XQYojlaD/pubhtml
The model is meant to be used to predict the preference scores of different speakers, though, not a single speaker with and without EQ. There's a lengthy and on-going discussion about the optimal approach to EQing: direct sound vs PIR. I've listened to the arguments from both sides and decided I'll go back to a LW approach, this time with a -0.2 dB/oct. slope. If a resulting EQ profile does not improve all score components (excluding LFX), a PIR based replacement will be made.
Regardless of your stance in the debate, I can say with a high degree of confidence that my EQs improve the fidelity of your speakers, assuming low tolerances from the manufacturers.
It is pretty. Does not give any new information, but pretty nice for someone who is learning how to read different directivity plots. The only potential downside is that inexperienced viewer will look at all the plots one after another and not realize that it is exactly the same information over and over
It is really cool to support plots that are directly comparable with other important sources (for example SoundStage!) and otherwise focus on a few key plots that are easiest to read and compare.I am most happy with preference score rankings with Sub + EQ. That gives a good initial filter what is what for people who care to add Sub and will use either PQ or some automated stuff - Audissey, Dirac, ARC, etc. Then you can always go as deep into the data as you want, analyzing all the possible limitations and combinations.
Just my take on this.
OP
This plot also looks cool. If Understand it correctly, the middle is 20Hz and outer perimeter is 20kHz? Regular IsoBand plot still is more informative, but this one is more intuitive instead Just as Erin explained.
Just as Erin explained.
OP
- Thread Starter
- #70
New: the software now generates PEQs to optimize the speaker response based on @amirm measurements.
similar to what others are doing the software optimize for a LW which is as close as possible to a target and for a SP as flat as possible. It more or less optimize the pref score at the same time.
The optimizer works well and results for all speakers are in this directory on GitHub.
Well means that it gets similar results as what others are doing. @flipflop does it manually and is slightly better for now
Next step is to correct in-room measurement. Prototype is working and can do 2.0 and 2.1 with IIR, I am still work
ing on multiple subs. FIR will work soon or at least I hope so.
btw: if you have REW measurements of your speakers in-room: I would appreciate if you could send me
txt or wav export of l/r/subs with phase. I need a bunch to test the algorithm robustness.
similar to what others are doing the software optimize for a LW which is as close as possible to a target and for a SP as flat as possible. It more or less optimize the pref score at the same time.
The optimizer works well and results for all speakers are in this directory on GitHub.
Well means that it gets similar results as what others are doing. @flipflop does it manually and is slightly better for now
Next step is to correct in-room measurement. Prototype is working and can do 2.0 and 2.1 with IIR, I am still work
ing on multiple subs. FIR will work soon or at least I hope so.
btw: if you have REW measurements of your speakers in-room: I would appreciate if you could send me
txt or wav export of l/r/subs with phase. I need a bunch to test the algorithm robustness.
New: the software now generates PEQs to optimize the speaker response based on @amirm measurements.
similar to what others are doing the software optimize for a LW which is as close as possible to a target and for a SP as flat as possible. It more or less optimize the pref score at the same time.
The optimizer works well and results for all speakers are in this directory on GitHub.
Well means that it gets similar results as what others are doing. @flipflop does it manually and is slightly better for now
Next step is to correct in-room measurement. Prototype is working and can do 2.0 and 2.1 with IIR, I am still work
ing on multiple subs. FIR will work soon or at least I hope so.
btw: if you have REW measurements of your speakers in-room: I would appreciate if you could send me
txt or wav export of l/r/subs with phase. I need a bunch to test the algorithm robustness.
Thanks for this awesome benefit to the community.
Is there a way to reasonably interpolate the PEQ settings for people who have more limited systems? For example my Moode audio player supports 10 PEQ bands with a max Q of 8.0. So I cannot apply the recommended settings and am not sure the best way to approximate the recommended PEQ settings within my limits. Any guidance would be great.
posvibes
Senior Member
- Joined
- Jul 4, 2020
- Messages
- 362
- Likes
- 490
What a handy reference, great work!.
As I understand it the Kippel that Amir uses had two versions? A software update or something that increased the number of measurable points? Does that make a difference in results measured for speakers that were measured with the earlier version compared to those measured with the later version?
As I understand it the Kippel that Amir uses had two versions? A software update or something that increased the number of measurable points? Does that make a difference in results measured for speakers that were measured with the earlier version compared to those measured with the later version?
OP
- Thread Starter
- #73
hello,
you have 2 options: first is to take the filters in order (the first 10). second option is to use the software and generates the EQ you want:
Code:
./generate_peqs.py --helpwill give you the list of options. Here is an example with the JBL HDI-4500
Code:
./generate_peqs.py --speaker='JBL HDI-4500' --slope-listening-window=-0.5 --max-Q=8 --max-peq=10 --force
Code:
EQ for JBL HDI-4500 computed from ASR data
Preference Score 3.6 with EQ 5.3
Generated from http://github.com/pierreaubert/spinorama/generate_peqs.py v0.7
Dated: 2021-04-05-10:03:43
Preamp: -2.7 dB
Filter 1: ON PK Fc 546 Hz Gain +3.13 dB Q 3.41
Filter 2: ON PK Fc 14427 Hz Gain -4.34 dB Q 2.02
Filter 3: ON PK Fc 1914 Hz Gain +2.33 dB Q 5.49
Filter 4: ON PK Fc 10393 Hz Gain +2.66 dB Q 8.00
Filter 5: ON PK Fc 6120 Hz Gain -1.11 dB Q 1.00
Filter 6: ON PK Fc 1563 Hz Gain +1.17 dB Q 8.00
Filter 7: ON PK Fc 604 Hz Gain -1.19 dB Q 8.00
Filter 8: ON PK Fc 456 Hz Gain -0.78 dB Q 8.00
Filter 9: ON PK Fc 693 Hz Gain +1.04 dB Q 2.12
Filter 10: ON PK Fc 3002 Hz Gain +0.92 dB Q 8.00
OP
- Thread Starter
- #74
I think Amir always release data with the same number of points. Increasing the precision usually decrease the score. Data from vendors are usually more smoothed and that generates a higher preference score (from +0.5 to +0.7). I wanted to decreased the score automatically based on the amount of smoothing but I have not yet implemented it.
Thank you. I will try this!
devopsprodude
Senior Member
This is great stuff, thanks for doing this. Can you also please include a link back to the original review on ASR (and the other sites)? Thanks!
I don’t have much doubt. Though the Stereophile measurements show a bump around 1800Hz:https://www.stereophile.com/content/kef-reference-5-loudspeaker-measurements
The issue is the formula liking narrow directivity, with the Reference 5 is, so with real human trials it likely wouldn’t score as high as the formula predicts.
- Joined
- Jan 15, 2020
- Messages
- 6,864
- Likes
- 16,813
As discussed in many threads also not few people prefer narrow directivity, depending also on the room and listening distance, so such preference cannot be generalised.
Similar threads
- Poll
