SoundStage! Network NRC Measurements for Focus Audio

[microstrip]

I think there also has to be some context here. The distortion measurements are at 90dB with pink noise! Have you ever listened to pink noise at 90dB? It will run you out of the room. It is far more demanding than music at 85dB, which is what many will consider average listening.

I am prepared to accept that measurements should be taken in some context. But your argument to devaluate the meaning of the distortion measurement is not valid. Pink noise at 90 dB will drive you out of the room because of middle and high frequency energy, not bass.

As you properly said many excellent speakers have worst measurements in other areas. But their excellent sound does not disguise they measure badly. Numbers are numbers and no one can accept that 30% distortion at 60Hz is a good figure.

I own two speakers that for me sound great - the Soundlab's A1 Px and the Quad's ESL63. The first measure miserably and the second shows some of the best measurements ever recorded. But my main speakers are the Soundlab's - sometimes we believe in our heart, not in our brains...

 

[Jeff Fritz]

I am prepared to accept that measurements should be taken in some context. But your argument to devaluate the meaning of the distortion measurement is not valid. Pink noise at 90 dB will drive you out of the room because of middle and high frequency energy, not bass.

As you properly said many excellent speakers have worst measurements in other areas. But their excellent sound does not disguise they measure badly. Numbers are numbers and no one can accept that 30% distortion at 60Hz is a good figure.

I own two speakers that for me sound great - the Soundlab's A1 Px and the Quad's ESL63. The first measure miserably and the second shows some of the best measurements ever recorded. But my main speakers are the Soundlab's - sometimes we believe in our heart, not in our brains...

I think Doug addressed this better than me:
Distortion is measured at 2 meters, and plotted at 2 meters (we plot the regular at 1 meter). 90dB anechoic at 2 meters is 96dB at 1 meter, approximate 99dB in-room single speaker, 102 db in-room 2 speakers. For a 5" to play like that isn't bad at all.

No one said it was a "good figure," but you stated more comprehensively that the speaker showed "poor anechoic measurements," and it is with that broad statement that I disagree.

 

[microstrip]

No one said it was a "good figure," but you stated more comprehensively that the speaker showed "poor anechoic measurements," and it is with that broad statement that I disagree.

OK, it is a question of standards. I am not an audio engineer, but for me, taken overall, these measurements seemed poor. As usual, let's let the experts decide.

 

[Jeff Fritz]

Perhaps no one will send their speakers to be measured if % is used?

I don't think anyone is "hiding" in our measurements -- quite the contrary. We have the most revealing speaker measurements of all the NA magazines. Where else do you see distortion measurements at all? And honestly, if someone is dropping thousands of dollars on a speaker system, and they are concerned about the % verus dB, they could spend the two minutes it takes to convert them.

 

[RBFC]

Jeff, could you post a quick table that correlates %THD with dB Distortion? 10-20 increments of 1%THD would be sufficient for most applications here. You might include some fractional THD% also! Start a new thread in the General Audio Discussion forum and one of the moderators will make it a "Sticky" post for reference purposes.

Much appreciated,

Lee

 

[Jeff Fritz]

Jeff, could you post a quick table that correlates %THD with dB Distortion? 10-20 increments of 1%THD would be sufficient for most applications here. You might include some fractional THD% also! Start a new thread in the General Audio Discussion forum and one of the moderators will make it a "Sticky" post for reference purposes.

Much appreciated,

Lee

As Doug pointed out, the reason % does not work is that the speaker is not flat in FR. No speaker is. So the percentage of distortion is constantly varying with frequency, which constantly varies across the spectrum. % does work if we are talking about a specific frequency. But our measurements show distortion across a wide FR spectrum, as they should. So it's not just a matter of redrawing a graph with % instead of dB. It just doesn't work like that.

 

[amirm]

Jeff, I don't follow. Whenever I see two graphs with different metrics composited, the right hand axis has the values for the other graph. That is what confused me from the beginning. Overloading the one axis for both seems odd. You can just as easily have the distortion percentages on the right hand side and leave everything else the same. No?

 

[RBFC]

As Doug pointed out, the reason % does not work is that the speaker is not flat in FR. No speaker is. So the percentage of distortion is constantly varying with frequency, which constantly varies across the spectrum. % does work if we are talking about a specific frequency. But our measurements show distortion across a wide FR spectrum, as they should. So it's not just a matter of redrawing a graph with % instead of dB. It just doesn't work like that.

How about generating tables for a few specific "reference" frequencies? Say 60 Hz, 250 Hz, 1 kHz, and 10 kHz? Or, pick a few frequencies that you feel will best represent the speaker's capabilities across the spectrum. Then, when mentioning a certain speaker, we can see distortion figures at F1, F2, F3, F4, etc. for simple comparison purposes.

Lee

 

[Jeff Fritz]

Jeff, I don't follow. Whenever I see two graphs with different metrics composited, the right hand axis has the values for the other graph. That is what confused me from the beginning. Overloading the one axis for both seems odd. You can just as easily have the distortion percentages on the right hand side and leave everything else the same. No?

OK, I don't think I'm doing a very good job at this communication thing

I'll try again:

If the FR were a flat line, then you could have a graph with percentages because those percentages would correlate with the straight line. In this case, the distortion correlates with FR, which is wavy across the board. If there is a 6dB peak at 100Hz, for instance, and distortion is -30dB at 100Hz, then that "mark" on the chart is different than -30dB at 500Hz, where there might be a depression of 6dB. So even though both are -30dB relative to the FR, they are 12dB apart on the chart. Using decibels you can represent this on the graph because they "stand alone" as a value. A percentage does not stand alone. Make sense?

 

[amirm]

Jeff, are you saying it is impossible to have a graph which charts the distortion % in vertical axis, and frequencies in the horizontal axis? If so, then I am lost in the woods .

 

[Jeff Fritz]

Jeff, are you saying it is impossible to have a graph which charts the distortion % in vertical axis, and frequencies in the horizontal axis? If so, then I am lost in the woods .

Until I can figure out another way to explain it, here is the way to convert:

Both curves are reported in dB which can be read off the vertical axis. In order to convert to a percentage one must read the top line (frequency response) and then determine the dB difference between that line and the bottom line (THD+N line). Translation from dB to % is as follows:

Equal (or 0dB difference) = 100 %
-10dB = 31.6%
-20dB = 10.0%
-30dB = 3.16%
-40dB = 1.0%
-50dB = <0.5%

 

[Jeff Fritz]

Jeff, are you saying it is impossible to have a graph which charts the distortion % in vertical axis, and frequencies in the horizontal axis? If so, then I am lost in the woods .

If you had a speaker with 10% distortion across the board, and an FR that was not flat (all speakers), how would you draw the line on the vertical axis? 10% across the board would not be a straight line, but a line identical to the FR. So you couldn't draw a straight line to represent the distortion. So how could you have a % vertical axis?

Perhaps there is a way, but I'm not seeing it. Where is Sean Olive when we need him!

 

[microstrip]

Or use this online converter:

http://www.sengpielaudio.com/calculator-thd.htm

But we should remember loudspeaker distortion and electronic distortion are two distinct entities. One distinct member of this forum (Dr. Earl Geddes) has written some very interesting papers on it. See papers Distortion Perception I and Distortion Perception II at :

http://www.gedlee.com/distortion_perception.htm

 

[amirm]

Jeff, please help my understanding of the graph. The top line shows the amplitude of the signal produced by the speaker. The bottom line is the level of distortion generated as a result. Both are levels. Therefore, to find the percentage, I find the difference between them and that would tell me how low the distortion is relative to the signal. Then I apply the usual formulas to get my distortion %.

If so, I went ahead and performed the above math for the speaker in question and I get this table:

Frequency %distort THD_Diff (db)

50	   31.00	10
100	    5.01	26
200	    2.82	31
500	    0.45	47
1000	    0.45	47
2000        0.50	46
5000	    0.63	44
10000       0.32	50

Am I good so far? If so, then let's plot the above percentages:

Is this right?

 

[RBFC]

Not Jeff, but butting in.

Looks good to me, Amir!

The problem seems to be that we will need to generate a graph like this for different levels of playback loudness... not the end of the earth, but extra work. Would some type of 3D waterfall-like graph work?

It's certainly apparent where the work in audio reproduction still needs some focus, huh?

Lee

 

[Jeff Fritz]

Jeff, please help my understanding of the graph. The top line shows the amplitude of the signal produced by the speaker. The bottom line is the level of distortion generated as a result. Both are levels. Therefore, to find the percentage, I find the difference between them and that would tell me how low the distortion is relative to the signal. Then I apply the usual formulas to get my distortion %.

If so, I went ahead and performed the above math for the speaker in question and I get this table:

Frequency %distort THD_Diff (db)

50	   31.00	10
100	    5.01	26
200	    2.82	31
500	    0.45	47
1000	    0.45	47
2000        0.50	46
5000	    0.63	44
10000       0.32	50

Am I good so far? If so, then let's plot the above percentages:

Is this right?

No. Your graph shows linear distortion between the points. Our graph shows, for instance, no measurable distortion at 500Hz. Your graph smoothes this to the extent that we lose all resolution and the speaker seems to have linear distortion across the frequency band, which clearly isn't the case.

 

[FrantzM]

Or use this online converter:

http://www.sengpielaudio.com/calculator-thd.htm

But we should remember loudspeaker distortion and electronic distortion are two distinct entities. One distinct member of this forum (Dr. Earl Geddes) has written some very interesting papers on it. See papers Distortion Perception I and Distortion Perception II at :

http://www.gedlee.com/distortion_perception.htm

I also did find these papers very interesting .. We don't seem to perceive distortion (at least the way it is usually measured) very well... High level of THD sems perfectly totlerable ...

 

[das]

Hello,

Jeff pointed me to this forum and I'll weigh in on a few things.

1) As Jeff pointed out, you can't just % to the left-hand side because it varies with each point. The distortion is the relation of the input signal, top, to the distortion components. Decibels are used for the both, so it's the difference between the two lines.

2) People like to see distortion percentage, and that can be done with a different graph that would simply express that. On the other hand, you'd lose pertinent data, such as what the shape of the curves look like, which might give you more indicators as to what the cause and/or effect of that distortion might be. In short, you're best off to learn to read in dB (there's a reason it's used in engineering and scientific areas) and, also, learn more about interpreting graphs.

3) I find it odd that comments get made on the relative quality of the distortion measurements when, in fact, we're the only publication to produce distortion measurements. Stereophile doesn't. So what the reference? The FP50 speaker is a very small speaker with a 5.5" woofer. It can barely extend to 50Hz at appreciable out, therefore, the distorition is bound to be high there. It's at its limit -- that's really what this graph is showing. Also, 90dB at 2 meters in a chamber is extraordinarily loud. People read 90dB and forget the environment around. As I pointed out in a separate email, that's 96dB at 1 meter, 99dB in a real room (gain), and that's just one speaker. Far, far, far in excess of what most would play a tiny speaker like this at.

4) Finally, if you want to compare and see "good" and "bad" measurements, spend some time at www.speakermeasurements.com

Thanks,

Doug Schneider
www.SoundStageNetwork.com

 

[microstrip]

Amir,
You wanted it just as it was done in the good old days?
from Hifi News and Record Review October 1981

 

[Ron Party]

Doug, welcome to the Forum!