[john1970]

To all AC viewers,

Attached is a link to an interesting page detailing amplifier power requirements.  It takes into account all of the necessary parameters (distance from source, desired listening level, loudspeaker sensitivity, and amplifier headroom) and calculates the required amplifier power.

http://www.crownaudio.com/apps_htm/designtools/elect-pwr-req.htm

Thought it might be useful,

John

[kfr01]

That's a nice bookmark.  Geared toward pro-audio applications, but useful nevertheless. 

One question I always have when looking at these things:

How much headroom is enough for transients?  Especially in big, well-recorded classical recordings.

3db, 10db? 

I'd think 10 (thinking transients might be twice as loud).  Anyway, 10 helps me justify the 85lb space heater in the room.  

[zybar]

If I use 10db of headroom, I need 250 watts.

Not too much below the 300 watts/8 ohms I have with my BAT VK-600SE.

George

[john1970]

I believe that a transient that is twice as loud is +3 dB.  +10 dB is actually 10 times as loud!

To calculate use the following formula:

dB = 10* log (P2/P1), where P2 and P1 are output power and P2 > P1

Here is an interesting discussion of dB and loudness:

http://www.phys.unsw.edu.au/~jw/dB.html

Cheers,

John

[Brian Walsh]

I believe that a transient that is twice as loud is +3 dB.  +10 dB is actually 10 times as loud!

No, something 10 dB louder requires 10 times the input power but is just twice as loud. A 3 dB difference is perceptible but is not a large difference, while a 5 dB difference is clearly audible.

To calculate use the following formula:

dB = 10* log (P2/P1), where P2 and P1 are output power and P2 > P1

P2 doesn't have to be greater than P1, although of course for this example it is. The figure you are calculating by that formula is the dB difference. If P2 < P1 then the dB difference is negative.
 

[zybar]

I believe that a transient that is twice as loud is +3 dB.  +10 dB is actually 10 times as loud!

john

I don't think so.   

George

[kfr01]

I believe that a transient that is twice as loud is +3 dB.  +10 dB is actually 10 times as loud!

It takes twice as much power to obtain a 3db rise in spl.  Our ears hear a 10db rise as being "twice as loud."

[john1970]

I stand corrected.  Thank you.

John

[kfr01]

I remember reading somewhere that Bob Carver used to recommend upwards of 1,000w to handle transients.

While I think that might be excessive, I wonder if there are any resources out there talking about the subject.

How much headroom is needed to handle transients in certain music? 

Anyone?

[JohnR]

According to this page, you want 50 dB peak to average for orchestral music.

http://www.audioheritage.org/html/perspectives/drews-clues/1-intro.html

I'm inclined to believe it, as to my ears most systems are unable to reproduce transients like "real" sounds. Just drop a glass on the kitchen floor, or drop a metallic object onto concrete, and then ask yourself if your system could reproduce that. The same thing for real instruments, esp brass, percussion and the like.

[fredgarvin]

thanks for the link John, how much power do you need for those RM40's according to the calculator?

[R_burke]

I may be showing/proving my ignorance here  but what is a normal listening range in terms of dBSPL.  I listen on the high side, but not cranking

[Russell Dawkins]

According to this page, you want 50 dB peak to average for orchestral music.

http://www.audioheritage.org/html/perspectives/drews-clues/1-intro.html

I'm inclined to believe it, as to my ears most systems are unable to reproduce transients like "real" sounds. Just drop a glass on the kitchen floor, or drop a metallic object onto concrete, and then ask yourself if your system could reproduce that. The same thing for real instruments, esp brass, percussion and the like.

thought provoking link, John.

I think that the provision for adequate dynamic reserves, both in amplifiers and speakers, may be an important (unsuspected) factor in fidelity of playback. To this end, it would be useful if amplifier makers listed in their specs the undistorted (< 1%) power available for 20 msec, 10 msec, 5 msec and 1 msec. This way you could get a picture of how the amp will behave on these short transient peaks. I suspect that when playing loud-ish we are often listening to how an amp sounds in the clipping-recovery mode, without realizing it.

What complicates the process is the (too) slow response time of most level measuring devices which simply do not register or display these very short transients.

[fredgarvin]

I usually listen at about 74db. WhenI want loud I might go to 80-85 but never with orchestral music. At 74 you can feel it slammimg your chest.

[BobRex]

I seem to recall that most CDs are "cut" at 20dB peak to average.  (perhaps Steve could confirm) That sounds like a more realistic figure than 50dB.  Considering that Pearson once measured 96(?)dB peaks in Carnegie, a 50 dB peak to mean would place the mean at 46dB??  Even giving 10dB for error, that's still 56dB mean.  That sounds low.  Given typical compression, the 20dB figure sounds much more realistic. 

Now..., if you listen at 75dB average, that means you need to generate 95dB peak.  Calculate what you need to hit that, plus a little head room and you'll know how much power you "really" need.  Now qualify that with impedance dips and low freq. power requirements and you might be suprised at what you can honestly "get away with".  Hint: You don't need 1000 watts, possibly not even 100 

[andyr]

That's a nice bookmark.  Geared toward pro-audio applications, but useful nevertheless. 

One question I always have when looking at these things:

How much headroom is enough for transients?  Especially in big, well-recorded classical recordings.

3db, 10db? 

I'd think 10 (thinking transients might be twice as loud).  Anyway, 10 helps me justify the 85lb space heater in the room.  

This is a verry inneresting issue started by John1970 and I think you've hit the nail on the head, kfr01 ... what is the overhead required to cope with transients without clipping!!??     I suspect that is the magic number which transforms an average listening experience into a spine-tingling one!!

Playing with the Crown spreadsheet in John's link (3.5m, 83dB sensitivity of my Maggie IIIAs and 80dB SPL at the listening position, 10dB of headroom requires a lousy 60w but 20dB requires 600w!     30dB requires 6,000 watts!!

And if 10dB is only twice as loud ... how many dB does it require to be 10 times as loud??     I suspect 10 times is not out of the question for the millisecond leading edge of a cymbal??

I actually run my Maggies 3-way active with (all ratings are into 4 ohms):
* 220w on the bass panels
* 110w on the mid panels, and
* 50w on the ribbons.

Now, the "rule of thumb" with actives is you can double the effective power delivery, compared to a single amplifier driving a passive crossover - so I have about 750wpc available to me.  I certainly have never ever heard the slightest bit of clipping ... and true-ribbon Maggies have a very effective indicator of clipping - you tend to blow your ribbon!     And my cheap digital SPL meter often measures 80dB (when I can be bothered to take it out!) ... although, again, when people use their SPL meters ... do they have them on the correct settings (which I understand are: "slow" and "C-weighted")??

Fred, how did you have your meter set?

So, the key point to me is ... what headroom should we allow for transients?  I'm inclined to believe what JohnR posted - that to realistically reproduce the sound of a glass object breaking on a hard floor, a huge amount of headroom is required.

So it would be really helpful if, as Russell has posted, the bloody amp mfrs gave their undistorted power delivery specs into 20, 10, 5 and 1ms.  Then we could see what sort of headroom their power supplies delivered!   

Now, as I understand it, as BobRex has posted, CDs are typically cut (for marketing reasons) at a 20dB peak-to-average difference. However, LPs don't suffer from this compression ... so probably LPs deliver far more peak-to-average SPLs than CDs?

Regards,

Andy

[Brian Walsh]

And if 10dB is only twice as loud ... how many dB does it require to be 10 times as loud??

Let's see. If 10 dB greater is subjectively twice as loud, you can express the relationship mathematically as

2^(dB_diff/10) = subjective loudness ratio

such that if dB_diff is 10, 2^(10/10) = 2^1 = 2

Rearranging the first equation by taking the log of both sides first,

dB_diff = log(loudness_ratio) / log(2) * 10

So if subjective loudness ratio is 10,

dB_diff = log(10) / log(2) * 10 = 33.2

Going back to the original power ratio equation,

10 * log(P2/P1) = dB_diff

10 * log(P2/P1) = 33.2

P2/P1 = 10^3.32 = 2099

So to produce a sound ten times as loud requires about 2100 times as much power - assuming everything is linear and there are no losses, which in the real world isn't the case.

[andyr]

And if 10dB is only twice as loud ... how many dB does it require to be 10 times as loud??

Let's see. If 10 dB greater is subjectively twice as loud, you can express the relationship mathematically as

2^(dB_diff/10) = subjective loudness ratio

such that if dB_diff is 10, 2^(10/10) = 2^1 = 2

Rearranging the first equation by taking the log of both sides first,

db_diff = log(loudness_ratio) / log(2) * 10

So if subjective loudness ratio is 10,

db_diff = log(10) / log(2) * 10 = 33.2

So to produce a sound ten times as loud requires about 33 times as much power.

Thanks for doing the maths, Brian ... but but but!!   

OK, to produce a sound 10 times as loud requires 33 times the power.  That I can understand!

But what is the base-power level which we multiply by 33?

If you converted the "33 times" back to a "dB of headroom" figure for that Crown spreadsheet, we could crunch some numbers!   

Regards,

Andy

[Brian Walsh]

I posted too soon, andyr...please see my edited post. Please also note that log base 10 is used, not natural log base e.

As for base power level, it can be whatever you want. Conversely, you can back solve for it based upon loudness ratio and maximum power available.

[andyr]

I posted too soon, andyr...please see my edited post. Please also note that log base 10 is used, not natural log base e.

As for base power level, it can be whatever you want. Conversely, you can back solve for it based upon loudness ratio and maximum power available.

Hi Brian,

Jeez, you must be up early!!   

OK, I read your edited post ... to produce sound 10 times as loud requires a whopping 2,100 times the power!        Whew!!   

If I use the following numbers in that Crown s/sheet:
* 3.5m distance
* 80dB average level
* 83dB speaker sensitivity
* 0dB headroom

I get 6w.  Obviously this is a steady-state reading.

But 2,100 times this power to cope with transients which are 10 times as loud ... sheeech!!         I think my assumptions are wrong somewhere!  Mebbe a more "real world" figure is only 5 times?

What multiplying factor does that produce?

Regards,

Andy