Master Preference Ratings for Loudspeakers

[MZKM]

Looks like the charts are now visible on iOS w/ Safari!

I noticed that Google updated Sheets and added some features (e.g., hyperlink using select text and not the whole cell); and being hopeful, I checked and it looks like it now works (formatting is still a bit wonky though).

If anyone has an iPhone/iPad and can confirm on their end, that’d be great.

 

[napilopez]

@MZKM @edechamps, how hard would it be for me to try to compute a score with my measurements? I don't plan on using them to compare with Amirs or create my own ranking, but it could be a good data point for me to consider in my evaluations, since occasionally there are qualities that are captured by the speaker. There are a few speakers I've measured that overlap with Amir's so I'm curious how the scores would compare.

 

[MZKM]

@MZKM @edechamps, how hard would it be for me to try to compute a score with my measurements? I don't plan on using them to compare with Amirs or create my own ranking, but it could be a good data point for me to consider in my evaluations, since occasionally there are qualities that are captured by the speaker. There are a few speakers I've measured that overlap with Amir's so I'm curious how the scores would compare.

As long as you have the data points for the:
• on-axis
• listening window
• sound power
• PIR
They just get pasted in and it auto-calculates for mine. Now, it currently is setup to read it how Amir publishes it, but if yours is a bit different it shouldn’t matter, just some tweaking.

 

[napilopez]

As long as you have the data points for the:
• on-axis
• listening window
• sound power
• PIR
They just get pasted in and it auto-calculates for mine. Now, it currently is setup to read it how Amir publishes it, but if yours is a bit different it shouldn’t matter, just some tweaking.

Cheers! I took a look at the master file you'd shared a while back and couldn't get it to work right, which seems to be because REW exports too many points per octave by default(96). It doesn't let me do 20 like I think amir does, but it worked fine with 12.

Anyway, I tried the spreadsheet with the Amphion I recently measured since eyeballing it, it seemed to measure pretty similar to the Elac. The elac is a speaker I didn't think looked that great when I saw Amir's review but measured better than I expected, so I'm trying to better correlate my assessments with the score. For reference:

Amphion:

Elac:

Long story short, got a 5.3/7.3, which isn't too far off.

Granted, the lower resolution of quasi-anechoic measurements may be helping, although it also tends to cause squiggles in REW that aren't present in the anechoic data. Going to try some of the speakers I've overlapped with Amir now, though those are based off incomplete spin data (I used to only measure in the front hemisphere).

 

[MZKM]

Cheers! I took a look at the master file you'd shared a while back and couldn't get it to work right, which seems to be because REW exports too many points per octave by default(96). It doesn't let me do 20 like I think amir does, but it worked fine with 12.

Anyway, I tried the spreadsheet with the Amphion I recently measured since eyeballing it, it seemed to measure pretty similar to the Elac. The elac is a speaker I didn't think looked that great when I saw Amir's review but measured better than I expected, so I'm trying to better correlate my assessments with the score. For reference:

Amphion:
View attachment 62734

Elac:
View attachment 62735

Long story short, got a 5.3/7.3, which isn't too far off.

View attachment 62736

Granted, the lower resolution of quasi-anechoic measurements may be helping, although it also tends to cause squiggles in REW that aren't present in the anechoic data. Going to try some of the speakers I've overlapped with Amir now, though those are based off incomplete spin data (I used to only measure in the front hemisphere).

I am not 100% if I made changes or not since I shared that document, here is a copy of the current version:
https://docs.google.com/spreadsheets/d/16OX21-GxgiduKP2zEp4CV7EQDWz-tv-828rdK7LYzXs/copy
Do note that from the first speaker till now, the Klippel may have been spitting out the wrong PIR curve, as when I tested it against me manually calculating it, the few speakers I tried came out with a 0.1 point difference. Amir has contacted Klippel to see if it is indeed a bug. Not a huge deal, but it is something to note.

 

[napilopez]

Thanks! And yes I noticed that. There's also the matter of the early reflections curve discussed a while back, which will affect the PIR too. AFAIK Klippel hasn't addressed that, though I know Amir reached out to them. Someone mentioned there's a klippel update coming soon so maybe we'll see some of those changes then.

 

[MZKM]

Thanks! And yes I noticed that. There's also the matter of the early reflections curve discussed a while back, which will affect the PIR too. AFAIK Klippel hasn't addressed that, though I know Amir reached out to them. Someone mentioned there's a klippel update coming soon so maybe we'll see some of those changes then.

I think Amir has stated he has manually added a fix. I noticed the PIR was different when manually calculating but I don’t recall seeing the ER being different, and as I used a recent speaker to test this, the manual fix seems to be working.

 

[amirm]

Thanks! And yes I noticed that. There's also the matter of the early reflections curve discussed a while back, which will affect the PIR too. AFAIK Klippel hasn't addressed that, though I know Amir reached out to them.

They did actually. Unfortunately they did not give me a patch for the software. But rather, a back-door to manually apply that fix on every measurement which is time consuming so i have not used it. I am hoping they put out an update with all the fixes.

 

[napilopez]

L200 got a 5.5/7.5

They did actually. Unfortunately they did not give me a patch for the software. But rather, a back-door to manually apply that fix on every measurement which is time consuming so i have not used it. I am hoping they put out an update with all the fixes.

Oh, sucks that it's complicated, but I'm glad they addressed it - hopefully they do that soon.

 

[napilopez]

Managed to get it working for 24 pts/octave. There's a major caveat in that some of my older measurements (the S400 and R3) are based off incomplete measurements, but I was curious nonetheless.

For the speakers I've plugged into the spreadsheet so far (it's a pretty slow process because the data needs to go through various conversions before it works right), there seems to be quite a tight distribution:

Amphion Argon1: 5.3/7.3
Q Acoustics 3030i: 5.2/7.5
Polk L200: 5.6/7.6
KEF R3 (incomplete data): 6.3/7.8
Buchardt S400: 6.3/8.1

The S400 And R3 seem to have swapped places a compared to Amir's data, but in my book that's a good thing as in my own measurements the R3 does look a little rougher than the S400.

All these speakers come off looking pretty good though. Might again be the resolution, but moving from 12 to 24 points per octave generally helped the scores. I'll have try some worse speakers...

EDIT: Realized that my L200 calculation was only using 12 points instead of 24. Scores went up a bit from 5.5/7.5 to 5.6/7.6

 

[edechamps]

@MZKM @edechamps, how hard would it be for me to try to compute a score with my measurements?

You might want to try @pierre who's already set up for computing scores from a variety of input formats.

Right now Loudspeaker Explorer strongly assumes the input is in the format Amir publishes (i.e. the Klippel export format). I'm not necessarily opposed to the idea of supporting data in other sources and formats, but that's pretty far down my TODO list. I'm not even finished implementing basic score calculation to begin with…

 

[QMuse]

@MZKM @edechamps, how hard would it be for me to try to compute a score with my measurements?

Are you using REW export format? Because if you are than the scoring can be computed after EQ as well.

 

[MZKM]

I'm not even finished implementing basic score calculation to begin with…

Are you trying to fully calculate Smoothness? Because I did and then found out Sheets can do this on its own. Besides discrepancies with the bands and whatnot, I was able to do the calculations themselves in Sheets in a few hours. I imagine actually coding is going to take much longer .

 

[MZKM]

Are you using REW export format? Because if you are than the scoring can be computed after EQ as well.

Keep in mind one should only EQ the on-axis for this purpose.

 

[edechamps]

Are you trying to fully calculate Smoothness? Because I did and then found out Sheets can do this on its own. Besides discrepancies with the bands and whatnot, I was able to do the calculations themselves in Sheets in a few hours. I imagine actually coding is going to take much longer .

That's what I'm working on, yes. It's not hard - using Statsmodels I can compute Ordinary Least Squares in literally one line of code, and that gives me everything I need in a single call - slope (SL), intercept, and r² (which is SM by definition).

It's not a question of difficulty, it's a question of how much free time I have and other projects I also like to do things cleanly and carefully and learn stuff in the process, so I'm making relatively slow progress compared to someone who just wants to reach the goal as quickly as possible.

 

[QMuse]

Keep in mind one should only EQ the on-axis for this purpose.

How do you mean that, as filter would also affect LW, ER and SP? Are you saying that getting flat on-axis should be the EQ target?

 

[MZKM]

How do you mean that, as filter would also affect LW, ER and SP? Are you saying that getting flat on-axis should be the EQ target?

Only if the on-axis is changed do we know how it affects the other curves. If you EQ the LW/ER/SP, you have no clue how it affects the curves that make them up.

 

[QMuse]

Only if the on-axis is changed do we know how it affects the other curves. If you EQ the LW/ER/SP, you have no clue how it affects the curves that make them up.

I don't agree. Let's assume I EQ the on-axis for flat response. That same filter would change the LW/ER/SP and again I wouldn't have clue how it affected the curves that make them up.

As I will end up without knowing the effect on underlying LW/ER/SP curves whichever curve I choose to EQ, I may as well EQ for best score, and there I have better chance EQ-ing the ER (or PIR) than on-axis.

 

[edechamps]

Only if the on-axis is changed do we know how it affects the other curves. If you EQ the LW/ER/SP, you have no clue how it affects the curves that make them up.

I'm confused. EQ'ing a speaker is an all-or-nothing proposition: it will affect all curves at once, because EQ applies all radiated sound equally. From that perspective, I don't understand what you're trying to say by "If you EQ the LW/ER/SP, you have no clue how it affects the curves that make them up". It's not possible to "EQ the LW/ER/SP". You can only EQ the entire speaker, and whatever EQ you apply affects all curves equally in a perfectly predictable way.

 

[QMuse]

I'm confused. EQ'ing a speaker is an all-or-nothing proposition: it will affect all curves at once, because EQ applies all radiated sound equally. From that perspective, I don't understand what you're trying to say by "If you EQ the LW/ER/SP, you have no clue how it affects the curves that make them up". It's not possible to "EQ the LW/ER/SP". You can only EQ the entire speaker, and whatever EQ you apply affects all curves equally in a perfectly predictable way.

Exactly. And because of that my EQ strategy would be aiming for best score, and not for flat on-axis.