Ideas for new ways to display data

[shadyJ]

I think averages should just be avoided. Averaging in dB will weigh down the average to the lower value, and averaging in Pa will pull the numbers up. To get an idea of the sub's performance at a glance, the best thing to do is what you are already doing; using comparative graphs. There is just no great way to boil down the performance of a certain frequency band with a single number, but graphs can do it all with a single image. 

[SME]

I agree that doing an average like this is not a good idea.  There's no real fair way to do it.  Everyone would have his/her own criterion for weighting the data.  And in fact, those of us who would actually care about how that average was calculated would not find it useful anyway.  Such numbers are only likely to be "helpful" to the sort of people who start forum threads with "I have $500 to spend.  Which subwoofer should I buy?" and no other details.

[Kyle]

"There's no real fair way to do it" -- this is the truth. The passive systems with 10k+ amps wil always have an advantage in the upper bass where active systems are always power limited. I think we'll do some nice filters for the chart view so we can compare active vs active and passive vs passive if users want.

[Ricci]

Doing away with the burst averages entirely certainly would avoid a lot of the issues. Sounds good to me.

Should we get rid of the long term sweep averages as well? I'm thinking yes...

I never see anyone discussing those and when I do it's an email from someone who is confused and trying to compare it to 1m burst data from somewhere else. Taking them away solves the problem with minimal effort. The correct way is to pull up the full bandwidth measurement and compare and we already have that data capability built in.

 

Just to illustrate how confusing some of this is here is some of the M.A.U.L. data using the various reporting methods.

 

Normal 2m "rms" method for reporting THD limited burst SPL used here at DB.

 

20Hz = 128.6

25Hz = 130.6

31.5Hz=133.1

40Hz = 135.8

50Hz = 138.3

63Hz = 141.4

 

Of course the 1 meter "peak" results would be 9dB higher which is easy to understand at least.

20Hz = 137.6

25Hz = 139.6

31.5Hz=142.1

40Hz = 144.8

50Hz = 147.3

63Hz = 150.4

 

Example simple SPL averages using the 2m data normally used here. Of course the 1m peak results would be 9dB higher.

 20-31.5Hz=130.8 (139.8dB 1m peak)

40-63Hz = 138.5 (147.5dB 1m peak)

 

Here are the same results averaged in pascals.

 20-31.5Hz=131 (140dB 1m peak)

40-63Hz = 138.8 (147.8dB 1m peak)

 

We can see that the results using Pa averaging are just a hair higher in this case. They are always higher than averaging using SPL. It doesn't amount to much here because the M.A.U.L. is actually quite uniform in output over these ranges. Varying by 12.8dB from 20Hz to 63Hz which seems like a lot but it is actually not that much. With systems having a decently uniform output the Pa averaging doesn't boost the output averages that much so it isn't a big deal. Let's look at a system which is not as uniform in output though.

 

Here's the Bic PL-200 THD limited burst data

 

20Hz = 84.0 (93dB 1m peak)

25Hz = 96.2 (105.2dB 1m peak)

31.5Hz=102.5 (111.5dB 1m peak)

40Hz = 107.7 (116.7dB 1m peak)

50Hz = 110.4 (119.4dB 1m peak)

63Hz = 108.1 (117.1dB 1m peak)

 

Averages of above PL-200 data as presented here. 2m, rms, straight averaging of SPL

 20-31.5Hz= 94.2

40-63Hz = 108.7

 

Averages of the same data but reported at 1m, peak, with Pa averaging

 20-31.5Hz= 106.1

40-63Hz = 117.8

 

In this case the 20-31.5Hz average is greatly boosted by 2.8dB by using Pa averaging. The PL-200 does have a wide variance over that bandwidth but it is by no means alone or the worst offender. There are many systems that have been tested that are similar. I believe that everywhere else reporting averages or most everywhere else anyway are using this Pa averaging method. It is better to simply look at the discrete frequencies individually rather than lump them together as averages in those cases IMHO. If you compare to another system that has much more uniform output down to deeper frequencies but do not look at the discrete bandwidths and only the averages when comparing systems, it would appear that the less uniform system is more comparable in deep bass output than it actually is.

 

Does anyone have a counterpoint or think we should keep any averages at all after the next big update?

[Kyle]

I agree because its also less work for me with the new site.

[Kyle]

The new site is going to have a lot more rich experience anyway, and the comparing tools will be much better. I'm also thinking about adding a quick graph mode where any user can just add plot points and it will save the data to a hash key in the db and you can pass the link around, or edit it further similar to https://jsfiddle.net/, but with graphs, not code.

[lukeamdman]

What I liked about the 2m RMS value was that it was a little "closer" to reality, especially below 30hz when you factor in room gain.  Those numbers were actually reachable.  Add 9db and they're way out of reach. 

[Ricci]

What I liked about the 2m RMS value was that it was a little "closer" to reality, especially below 30hz when you factor in room gain.  Those numbers were actually reachable.  Add 9db and they're way out of reach. 

 

Yep. I have some data for powered subs in my current room vs outdoors and it matches up well. You see some wild variation up near 80-125Hz as you would expect from a single sub subject to the acoustics of a typical domestic sized room. The output tracks quite well overall from 25-70Hz and starts showing some gain at 20Hz (3dB roughly)and it increases steadily down to 10Hz and below (10dB or more). About what should be expected in general.

 

That leads to a good question. Since we are tossing out output averages...Should we entertain the idea of listing the burst data as both the standard way and the +9dB, 1m peak method? I'm thinking we don't and we perhaps just put a banner or note on the data that states that it is reported as 2 meter rms (Add 9dB for 1m peak equivalency) or something to that effect.

 

We will be listing the distortion limited burst data as usual and adding the maximum burst output with no regard for THD which I have for most systems. I'll usually run the system until it stops getting louder, THD reaches 100%, or I hear some sort of "BAD" noise indicating self destruction is nigh. Generally the extra output is in the deep bass frequencies below the system impedance maximum as can be seen on the static burst data images. The data is useful from my perspective as it shows what the actual dynamic maximum of the system is. Some of the higher inductance drivers for example fail in the deep bass for distortion a good 6dB or more below their useful excursion limits. For those drivers the distortion limited results aren't really a good indicator of how much displacement the driver truly has. When you place a sub in a room as compared to outdoors the THD often drops dramatically in the deep bass. Some powered subs have gone from being distortion limited during the 12.5 or 16Hz burst, a good 8dB below their maximum output outdoors, to producing a "pass" distortion result at very nearly full output in my room.

[lukeamdman]

That leads to a good question. Since we are tossing out output averages...Should we entertain the idea of listing the burst data as both the standard way and the +9dB, 1m peak method? I'm thinking we don't and we perhaps just put a banner or note on the data that states that it is reported as 2 meter rms (Add 9dB for 1m peak equivalency) or something to that effect.

 

YES. 

[Kyle]

YES. 

 

We can just have a toggle button that adds 9dB to each value in the table. Two different views of the same data.

[Kyle]

I'm doing away with long term and short term averages unless there are any objections with good arguments. We'll simply maintain 2m burst going forward for CEA2010 charts.

 

These changes will apply to dbv2

[Ricci]

No objections here. Less work for both of us and less confusion to casual viewers anyway.

[Funk Audio]

If there are no more long term averages shown what about putting the long term output in a chart just like burst output, just list the output at fixed frquencies, the same frequencies as burst and sort it the same way. Also it might be nice to even have that available to show on the same comparison graph as burst, so one can compare the same system, for burst and long term, as well as to other systems.

[Ricci]

Anybody want to vote on Mr. Funk's idea?

 

I'm not sure how I feel about that yet. Need to mull it over a little.

[shadyJ]

I, for one, like Mr. Funk's idea. The long-term output sweeps are often overlooked in favor of the burst output measurements, but they still do show an important aspect of the subwoofer's performance. I guess they are overlooked because they are not summed up in a single number as CEA-2010 does. 

[Kyle]

We're going to add sound pressure per liter in the new version. This won't be a logarithmic Y-axis scale, it will just be a straight up sound pressure (Pa) per liter because the scales for both are the same. e.g.a two fold increase in Pa = 6dB. So if a system is twice the size of another system and it makes 6dB more output then that system has the same Pa/Liter as the smaller system.

 

 

We could also look at sound intensity per liter but this is a little different. A two fold increase in sound intensity is only a 3dB gain so I'm not sure if this will favor the larger systems in a misleading way.  I tend to feel that Pa/Liter is the best metric.