Sensitivity, Efficiency, and Room Gain

[Deang]

Since we are all somewhat locked into these recurring discussions involving power, I thought I would share some info in an area I've been thinking about on and off over the last month or so.

I will start by saying that unless you have a properly calibrated SPL meter, any attempt at simply using the published sensitivity specification, and then adding "watts" to come up with a SPL, that is then applied in the context of "loudness" at the listening position -- is completely useless.

Those using the Radio Shack SPL meter fair much better, however, this only gives one an approximation, as it is not a "professional" measuring device. Also, those using the meter to plot a frequency response curve of the room/speakers, should be using the correction values for the meter. http://www.danmarx.org/audioinnovation/rsmeter.html

When someone says, "I usually listen around 85db", I think it's important that we start following up with some questions related to how the number was arrived at. Many use something like the following:

My speakers are 102db/w/m and I have 22 wpc.

.0157....watt......84db at 1m

.0313....watt......87db

.0625....watt......90db

.125......watt......93db

.25........watt......96db

.5..........watt......99db

1...........watt....102db

2...........watt....105db

4...........watts...108db

8...........watts...111db

16.........watts...114db

Most here know it doesn't always really work this way. There is room gain (+3b), and there are two speakers (+3), so 6db need to be added. Also, 3db then needs to be subtracted for every three feet one moves back from the speakers. It just so happens that because I only set 9 feet back, after making the adjustments, the chart works for me "as is". So, I should only need 1/60 of a watt to reach 85db, and 16 watts to reach an ear scorching 114db. What is really interesting here, is that I should only need 1 watt to reach what I consider to be a very, very loud 102db.

This isn't happening.

Using my "toy" meter, set to the C weighting, and put on top of my listening chair (without me in it, as reflections from my body would skew the readings), I can at the most reach peaks of 105db without discnernable clippling (characterized by pinching and excessive harshness). Using the "slow" response setting, the nominal SPL is close to 100db, but not quite. I consider this to be a substantial difference between what the chart says I should have, and what I'm actually getting.

Possible explanations:

1) The RF-7's aren't as sensitive as the indicated rating

2) My amp doesn't put out 22 watts

3) The meter is really "off"

4) The chart is "wrong"

Anyone feel like taking a shot at helping me figure out where almost 15db of my output is going? This has been the case with all my set-ups, regardless of the power rating of the amp -- it always seems to take much more power than it should to reach the 100db barrier.

You see the problem here. If someone says they listen primarily at 85db (using Klipsch), do we say they only need 1/60 of a watt, or 2 watts? And if someone says they like to go to 11 -- do we say they can get by with 8 watts, or 30?

[tillmbil]

Great question Dean, but you are supposed to have an answer for this question. Are you holding? I cannot get much over 110db's with my 79 CW's and 35wpc with the Dynaco. I should be able to at least hit 113db.

I thought it was watts to the 2nd power though. Like :

1

2

4

16

144 for each 3db's. I must be wrong huh?

[Deang]

I think I know, but I don't want to show my cards and risk looking like an ***. Besides, I want to wait for the heavy hitters to weigh in first.

A doubling of power gives a 3db increase, and it takes a ten-fold increase in power for a doubling of SPL.

[Deang]

Well now, that's the rub isn't it? Why of course, I'm using MUSIC. I don't listen to test tones being fed into a speaker with 2.83 volts. And of course, the whole deal is fluctuating with the impedance which impacts my total acoustic output.

[Painful Reality]

I would also try to get a more accurate estimation of clipping than your aural evaluation.

[WMcD]

Hmm,

Hard to tell exactly what you're doing.

Let me suggest that you start with a known power into the speaker with a test tone. You'll have to have an AC volt meter connected to the amp output. The normal reference is 2.82 volts to give a nominal 1.0 watt. If you put the RS meter 39 inches (1 meter) away, you should read the rated number for 'efficency'. I've done it and it comes out that way.

This assumes you have a CD with continuous test tones, a computer program for generating them, or a signal generator.

It is going to be very loud at 2.82 volts. 0.89 volts of drive will get the rated effency less 10 dB. 0.282 volts will be 20 dB down. This last is tolerable to listen to, but still quite a lot.

It is good to start with 1000 Hz.

The general rule is that on most music, peaks are 18 dB greater than the average. Call it 20 dB. This means there is 100 times the power on peaks than the average.

Also. I think you may be overestimating the gains from the room and the two speakers. Room gain is only in the bass. The two speaker gain only occurs when they are fed with identical signals, in phase. This doesn't happen in music.

Best,

Gil

[WMcD]

Hmm,

Hard to tell exactly what you're doing.

Let me suggest that you start with a known power into the speaker with a test tone. You'll have to have an AC volt meter connected to the amp output. The normal reference is 2.82 volts to give a nominal 1.0 watt. If you put the RS meter 39 inches (1 meter) away, you should read the rated number for 'efficency'. I've done it and it comes out that way.

This assumes you have a CD with continuous test tones, a computer program for generating them, or a signal generator.

It is going to be very loud at 2.82 volts. 0.89 volts of drive will get the rated effency less 10 dB. 0.282 volts will be 20 dB down. This last is tolerable to listen to, but still quite a lot.

It is good to start with 1000 Hz.

The general rule is that on most music, peaks are 18 dB greater than the average. Call it 20 dB. This means there is 100 times the power on peaks than the average.

Also. I think you may be overestimating the gains from the room and the two speakers. Room gain is only in the bass. The two speaker gain only occurs when they are fed with identical signals, in phase. This doesn't happen in music.

Best,

Gil

[JohnA]

"3db then needs to be subtracted for every three feet one moves back from the speakers."

Dean, this is one of your errors. SPL drops by 6dB with each doubling of distance, IIRC.

The other is that your Rat Shack meter (and mine) cannot HOPE to display musical peaks that are just a few dozen milliseconds long. It is not unbelieveable to me that there are peaks 10 dB above the RS meter peak, since it is so "slow", even on the fast setting.

[Deang]

"I think you may be overestimating the gains from the room and the two speakers. Room gain is only in the bass. The two speaker gain only occurs when they are fed with identical signals, in phase. This doesn't happen in music."

Gil, I don't understand. If I turn on one speaker, and then add the other -- it sure sounds louder. I didn't know room gain applied only to the bass. Thanks.

John, I was hoping there was a problem with my math.

[sfogg]

Gil,

"The two speaker gain only occurs when they are fed with identical signals, in phase.  This doesn't happen in music."

Sure it does, if you are sitting between your speakers and have a central image that is from material that is identical in both channels. Also bass tends to be fairly correlated as well.

Shawn

[sfogg]

John,

" SPL drops by 6dB with each doubling of distance"

Yes, for an omni-directional radiator. For a true line source though it drops off at 3dB for doubling of distance as the sound is more focused.

Shawn

[WMcD]

I was in error. You should get 3 dB with two speakers.

I was thinking about the situation where two closely spaced speakers fed with the same signal can give 6 dB, in theory.

Sorry.

Still, I believe you have to measure all these things for yourself in your room. All the theory is good in itself. Whatever the error of an RS meter, you should try a test tone at one meter, and then back farther into your room. You'll probably find some standing wave nulls and peaks.

It is easy to spout off numbers and theory (which I do often enough). When you do it for yourself, make measurements, keep notes, you gain insights and become a real world acoustic engineer.

Best,

Gil

[Klipschfoot]

Doesn't "room gain" correlate with room resonance? And, if so, how do you measure resonance? Then how does resonance affect sound realism? After, isn't realism what we are after?

[Deang]

You have to it Gil's way to learn about those things. A good test CD or signal generator, graph paper, a ruler, and a lot of patience.

Jeff, I think clipping is pretty easy to hear -- it goes from sounding very good, to sounding very bad.

My original point in all of this is that one should not rely on the sensitivity rating of the speaker and the charts as an indication what their speakers are doing in their room with their amp. We often ask, "how loud do you like to listen" -- and numbers are thrown out, and we don't give much thought to how the number is being generated. When I say I listen at 90-95db, it is a metered measurment using music at the listening position. This is very different than what I think some are doing.

[maxg]

Hmmm - its a mystery - the whole measurements thing - at least to me.

My take is that the measurements supplied are only really useful in making relative comparisons between speakers and not for establishing how loud they will actually play in your listening environment. Even then it is a bit touch and go.

To this end I would love to be able to measure the sentivity of my own speakers in the same way as they do for others. These are definitely less sensitive than the Heresy's (at 96 dB supposedly) but how much is impossible for me to tell.

If we are talking about measuring music volume (which I agree with others here is totally different from test tones) then the interaction of amp and speakers is critical. To give a few examples of what I am trying to say:

When we tested out a pair of RF3's on my old KT88 based 45 wpc amps they went louder than anything I have ever heard. I have no idea how loud - I didnt have a meter to hand then, but it was enough to send me running into the garden with my hands over my ears (whilst my friend Tony said - "That's loud" and reduced the volume.

Playing my Sansui 2500's which are rated to exactly the same 98 db/w/m the volume never got to anything like that level, and I mean anything like it!!

Playing my Heresy's (96 db/w/m) should have got close but never did. Relatively speaking I would say they appeared to be about 1/4 or less of the absolute volume of the RF3's.

The Sansui's did play slightly louder than the Heresy's. The Heresy's slightly louder than my home made speakers.

When I switched out the 45 wpc KT88's for 70 wpc EL34's I failed to get any increases in absolute volume. In fact, if anything, I have less. Despite this, it appears to be louder. In other words the SPL meter is giving very similar readings to those I had before but it feels louder. No idea why?

I get different results with different sources too. The CD player starts louder than the TT but doesnt go as high. The DVD player is quieter than both on some recordings and louder than both on others (MJ - History, for example is quiet, Prince Hits collection is the loudest recording I have).

Using MJ History the loudest volume I can get - even at peak, is about 99 dB at my listening position. Prince goes maybe 8-10 dB higher at peak.

Overall then my take is that there are so many variables going on that measurements are almost useless in this area.

I havent even covered the issue of changing impedance of the speakers and the loading effects that has on the amp.

Just out of interest I have 2 output taps on my amp - 4 ohm and 8 ohm. I thought the 4 ohm would be louder - it isnt. Another mystery to conjure with.

[djk]

A while back Bob G. was invovlved in a discussion of Klipsch sensitivity ratings.

Klipsch already adds 3~4dB for room gain to their published ratings, ie: an advertised 102dB is 98~99dB anechoic and with 2.83V/1M.

Many tube amps like these speakers on their 4 ohm taps, so the speaker is only 95~96dB/1W/1M.

10' is a good listening distance. (39.4"/120")log^10 X 20 = -9.67dB

22W = 13.42dBW

95~96dB -9.67dB = 86dB + 13.42dB = 99.42dB

Add 3dB for the other channel and you get 102.42dB peak before clipping sets in. Most people can't hear clipping in an amplifier only driven 3dB into clipping, so call it 105dB more or less.

"I can at the most reach peaks of 105db without discnernable clippling"

Sounds about right to me.

[Deang]

This is embarassing, but I guess I don't learn anything unless I ask.

(39.4"/120")log^10 X 20 = -9.67dB

What does the 20 represent in this formula and what does the "log ^" mean in the calculation?

"22W = 13.42dBW" -- how did you come up with this?

Neat stuff, I appreciate it.

[hurdy_gurdyman]

----------------

On 11/10/2003 5:21:00 AM djk wrote:

A while back Bob G. was invovlved in a discussion of Klipsch sensitivity ratings.

Klipsch already adds 3~4dB for room gain to their published ratings, ie: an advertised 102dB is 98~99dB anechoic and with 2.83V/1M.

Many tube amps like these speakers on their 4 ohm taps, so the speaker is only 95~96dB/1W/1M.

----------------

I am pretty sure my old Heresys are not really 96 dB as their rating would suggest. I own a pair of EV LS-12's which, according to EV, have a rating of 93 dB. Many on the Fullrange Driver Forum have measured them to actually be 95 dB. When I compare my Heresys with my EVs, the EVs are noticably louder at the same volume setting.

Also, I believe the older Heritage series speakers actually prefer the 16 ohm tap on many tube amps, probably because the horns have very high impedance through much of their range. My Heresys seem to be a bit louder on the 16 ohm tap of my Scott amp.

Dave

[Colin]

Please note gentlemen, that the one thing we have in common is the inexpensive RS analog SPL meter. Our rooms, loudspeakers, amplifiers, cables, interconnects, furniture's and test CDs are all different. Only the woefully inadequate RS SPL meter is the same. Yet, such were the tools of Galileo, Newton and Salk. They did NOT have modern equipment to make startling discoveries. We do NOT need the exact same equipment to make valid observations. Lets allow the RS meter about 2DB for errors and anomalies throughout most of its range and anything above 10K should really be taken with a grain of salt. Yet, despite the crudeness of the tool, we can still derive some very interesting relationships between rated loudspeaker efficiency and actual room loudness.

1. My vintage Dynaco Pas 2, series II, tube pre-amplifier by Panor Corp. supposedly does NOT have the typical volume control knob. On a typical pre-amplifier or receiver, the power knob gives out most of the output between 7 and 10 on the dial. I have measured this for myself. A typical potentiometer does NOT have a lot more gain after 3:00 on the dial. So 9:00 on the Pas 2 dial approximates ¼ of the amplifier output.

2. My audio/visual room is a carpeted 14 by 17' space with popcorn painted 8 high ceilings and concrete block construction on the front and right side-walls. When not stuffed with speakers, the space is a little bright.

3. Of the loudspeakers I have seriously auditioned, in my own home, with the same music and equipment, for EnjoyTheMusic.com, there are quite some sensitivity differences:

At 9:00 on the pre-amplifier dial, my super-sensitive walnut-oiled Cornwall 1s, with B2 crossovers, and rated at 100db/w/m, measured 68dB at the listening position, nine feet away, with me in the chair. This was at the slow C weighting. (In a much larger townhouse great room, they measured 65dB.)

The leaning Newtronics Skates transmission lines are rated at higher than average 88dD/w/m and showed 73dB with the same measurement, with my hefty 70s style vintage Class A solid-state Pioneer M-22.

Vince Christian's Urbane E6c Satellite System, rated at a very high 92.5db/w/m, measured 72dB in the same test.

Omega TS1 Single Driver Loudspeakers In Pearlescent Blue Arctic, rated at super-sensitivity of 96db/w/m, scored 68dB, with my flea-powered tube Bottlehead 2A3 Paramour amplifiers.

My big ole Klipsch classic corner horns rate an incredibly amazing 104Db/w/m sensitivity. Yet, when bi-amped with the vintage Pioneer M-22 and the tube Paramour amplifiers, they measured 85dB in the same test. A quick test just now with the M-22 gave me a 82db result.

4. Except for the big ole corner horns, there does NOT seem to be any strict relationship between rated loudspeaker sensitivity and their measured 1kHz frequency response, in my own home, with the same test CDs and equipment!

[sfogg]

Colin,

"there does NOT seem to be any strict relationship between rated loudspeaker sensitivity and their measured 1kHz frequency response, in my own home, with the same test CDs and equipment! "

You changed amps between each speaker test. Hence your testing is misleading as you didn't account for the differing amount of gain/input sensitivity in each amp. That you had the volume control on your pre-amp in the same position is meaningless as that is not the only thing that determines how much power you are putting into each speaker. Even using the same pre-amp and amp this method doesn't work since the impedance presented to the amps are all going to differ.

If you want to try measuring sensitivity you must measure the voltage going to the speaker then measure its SPL. The common test now is to apply 2.83 volts to the speaker on a 1kHz test tone then measure the SPL at 1 meter.

Shawn