[Roger A. Modjeski]

Well, gosh darn, this is a hot topic. The numbers are all over the place. I would love to hear from listeners who have actually measured the voltage at the speaker. That is what I suggested. That's what I want to hear about. That's all that really matters in this discussion. After that we might get into the problems some amplifiers have at clipping and below, one of them being intermodulation distortion.

Here's what I know. I have a nice Tektronix rack mount scope in my living room equipment rack that holds the rest of my system. I did this a few years ago because I had the extra scope, it looks cool and its always there. Anytime I think something funny is going on I turn it on and watch the music.  I especially watch it when I have a guest who likes it louder than I normally would play any kind of music. With small amplifiers I easily get average levels of over 90 dB SPL at the listening position about 12 feet from the speakers.  I often take it right up to clipping and let them listen for a bit then advance the volume just a few dB so there is occasional clipping. I've had good listeners be surprised to learn that  they did not hear clipping till it was happening about 10% of the time. Now keep in mind this is for my amplifiers... your amplifier may not recover nicely from clipping, many don't which may be a reason for the overwhelming support for really really big amplifiers.

[JerryLove]

Some of these numbers enter the realm of bovine excrement.  Think about it this way... Let's assume that a rock concert actually has max SPLs of 110dB.

In 1996 English House/Electronica band Leftfield performed at a measured 137db.

THX reference is 85db with 20db of headroom (so peaking at 105db). That's what you should hear in the center of a movie theater. (to set a reference most of us have experienced).

110db is reasonable, but it's not certain to include everyone. (also, as an aside: low-frequency can run up to 10db hotter)

And let's assume that you actually want to replicate that in your home (in my place it's not gonna happen).  Given a 90dB sensitivity speaker, that's 20dB or 100 watts.

...if you are 3' away.

Every time you double the distance, you lose from 3db to 6db. So at a listening distance of 12ft, you'd be -6db to -12db from what your speakers are outputting. So 400w-1600w on a 90db speaker.

That's within the dynamic envelope of most 50 watt amps (tube for sure, SS more than likely.)  Now here's where the "dynamic range" specification is stretched.  The 100dB dynamic range spec for CD is based more on its signal to noise ratio, and potential recording capability, than actual recorded signal content.  The number of recordings of any genre that are not compressed is miniscule, and in reality are more classical than anything else.

I have run measurements on much of my audio library. The good CDs are peaking 12-17db above measured average (I did have one hit 21db). The modern ones are often doing 6-9db peaks.

Show me a number of recordings that actually capture 100dB of dynamic range (thunder storms, trains, and jet planes excluded) and I'll agree with you, but until then these numbers just invite a circle jerk.

Technically anything that includes a silent moment and peaks at reference. But your point is valid.

[Roger A. Modjeski]

In 1996 English House/Electronica band Leftfield performed at a measured 137db.

THX reference is 85db with 20db of headroom (so peaking at 105db). That's what you should hear in the center of a movie theater. (to set a reference most of us have experienced).

110db is reasonable, but it's not certain to include everyone. (also, as an aside: low-frequency can run up to 10db hotter)
 ...if you are 3' away.

Every time you double the distance, you lose from 3db to 6db. So at a listening distance of 12ft, you'd be -6db to -12db from what your speakers are outputting. So 400w-1600w on a 90db speaker.
 
I have run measurements on much of my audio library. The good CDs are peaking 12-17db above measured average (I did have one hit 21db). The modern ones are often doing 6-9db peaks.
 Technically anything that includes a silent moment and peaks at reference. But your point is valid.

The THX reference of 85 dB with 20 dB headroom sounds quite reasonable for what they want to do. I'm surprised it isn't higher given that there are some very loud sound effects... re the scene in  DAS BOOT where the submarine and perhaps your subwoofer is breaking up. However I am not asking my home Hi Fi to do that. I am also not asking my home Hi Fi to replicate a rock concert where people who care about their hearing take earplugs. Can we please have a discussion about reasonable listening levels that we actually do 99% of the time. In the 1950's I am told that playing music "clean" at high levels was a bit of a contest. I also recall when I was working the sales floor at Audio Art in Richmond, VA that we always put the biggest amp we had (400 watts at the time) on the MG-2s and let them turn it up as loud as they wanted since we knew they couldn't hurt the speakers and they likely had never heard anything like that. Of course, I showed them the volume control and left the room posthaste!!!

Personally I find that with my ESL speakers and other speakers with good resolution that I don't need to play loud at all to hear ALL the music. By playing at lower levels I am able to listen longer and enjoy it more.

Although my interest had turned to small very clean amplifiers such as the OTL series I am currently working on and the direct drive ESL amplifiers that are now blowing away Acoustat owners who have never really heard their speakers before, I still make a 100 watt/ch amplifier and have sold far more of those than the smaller ones. The reason I opened this discussion was to see how many potential buyers are out there for smaller amplifiers. Although all audio equipment has gone through the roof in price, styling and complexity I retain my interest in bringing some very good sound to people at very reasonable prices with ease of service and high reliability.

[JLM]

Agreed Roger.  I tell others that the purpose of my dedicated listening room is to reduce background noise and thus be able to hear it all at lower spls.  As I mentioned above most audiofests I've attended keep averages at 80 dB and below (still higher than I normally do critical listening at).  But music is dynamic (made up of peaks) so they are ignored as your own risk.

Most don't know how loud they're listening habits are (or what frequency range they're listening to based on system capabilities - mostly bass, and treble - mostly hearing loss associated with abuse or aging).  It's easy to acclimate to louder/quieter listening (especially quieter, but I suspect based largely on individual personality) and not realize it.  It's just as easy to acclimate to loss of frequency range.  The human ability to adapt cuts both ways.  Years ago I took my speakers (rated 28 - 20,000 Hz, without the baffle step circuits   ) to an audio gathering of SET lovers and they were quite dumbfounded at first by hearing bass (set up well away from walls in a large open plan living room).  Note that the tube bloat compensated rather nicely for the lack of baffle step.  Like most others my age, I'm no longer 'bothered' by frequencies north of 15,000 Hz.  An audio salesman related to me years ago of a music professor customer from a major university who had been living for years with blown tweeters without realizing it, his mind filled in the blanks.  High fidelity reproduction simply requires that we fill in fewer blanks, making it easier to form the complete mental image of the performance (those with more musical experience/knowledge just need less help than the rest of us).

[Steven0100]

High fidelity reproduction simply requires that we fill in fewer blanks

That's interesting because this effect also helps people believe that their midfi system sounds as good as a high end system.  For instance, one person told me that my Quad ESL's sounded "almost" as good as his Bose.  I read somewhere where this store owner got so tired of hearing this that he set up Bose in his shop so people could A-B them together.  His sales shot up.

Anyway, like another poster mentioned, the more clarity and resolution you have, the lower volume we tend to play our music.  At least that's true in my case. 

As for THX, they can keep it, 85db average with 105 peaks is way too loud and not necessary.  That's one mans idea of perfection. 

I find that watching a movie at home is far more enjoyable compared to going to the theater.  At the very least, I don't have to give my hearing a break after the movie.  Besides too much realism is not good.  For instance, when there is a gun battle and hundreds of rounds are being fired, we would suffer hearing loss post-haste at realistic volumes. 

Most of my amps have been around 50 watts +/- 10 watts.  I've had amps at 130 watts, which is only about 3db louder.  Except for my 6 watt amp, I've never had an amp that couldn't exceed the volume that I wanted to listen at.  I suppose that the quality of amps makes a difference too.  Once I tried a Pass Labs 30 watt amp in one channel and a Carver 60 watt imp in another.  The Carver was peaking while the Pass Labs matched it in output with no distortion that I could hear. 

Personally, I think we are over thinking this. 

[BobRex]

In 1996 English House/Electronica band Leftfield performed at a measured 137db.

If that's the case, then everybody that attended that concert and did not use hearing protection, left with permanently damaged hearing.

Every time you double the distance, you lose from 3db to 6db. So at a listening distance of 12ft, you'd be -6db to -12db from what your speakers are outputting. So 400w-1600w on a 90db speaker.

Except that Rogers initial numbers, which is what I worked from, are based on both channels driven, thereby compensating for some of the distance roll off.  Also, those attenuation numbers are only for point source speakers.  Line sources and planars (again, Roger was working with his ESLs) drop off at less than half the rate.  So, combining those 2 factors, my numbers work fine.
 

I have run measurements on much of my audio library. The good CDs are peaking 12-17db above measured average (I did have one hit 21db). The modern ones are often doing 6-9db peaks.

Thank you for validating my original statement that CDs are cut with peaks 20dB above mean. 
 

Technically anything that includes a silent moment and peaks at reference. But your point is valid.

Yes, but anything that is recorded in a real room (as opposed to a anechoic chamber), by a microphone (not including pure electronic tones) will have silent passages restricted to the noise floor of the room.  Let's be generous and say that's 20dB, so the real dynamic range would be 80dB minus any compression or limiting, not the 100dB claimed.  That's why I call the number BS.

[JerryLove]

If that's the case, then everybody that attended that concert and did not use hearing protection, left with permanently damaged hearing.

I'm sure. I don't like normal concerts (and bring hearing protection on the rare events that I go)

Except that Rogers initial numbers, which is what I worked from, are based on both channels driven, thereby compensating for some of the distance roll off.  Also, those attenuation numbers are only for point source speakers.  Line sources and planars (again, Roger was working with his ESLs) drop off at less than half the rate.  So, combining those 2 factors, my numbers work fine.

Mono over both channels would improve efficiency by +3db (total power across both speakers) or +6db (power per-speaker). But I assumed there would be one speaker per-channel.

If you were trying to get 110db from a reasonable seating position: I'd recommend a speaker with a sensitivity higher than 90db.
 

Thank you for validating my original statement that CDs are cut with peaks 20dB above mean.

Only the good ones. I wish they were all so dynamic.
 

Yes, but anything that is recorded in a real room (as opposed to a anechoic chamber), by a microphone (not including pure electronic tones) will have silent passages restricted to the noise floor of the room.  Let's be generous and say that's 20dB, so the real dynamic range would be 80dB minus any compression or limiting, not the 100dB claimed.  That's why I call the number BS.

Actually: Redbook (16-bit) is 96.33db.

This isn't actually affected by noise-floor per-se, as one can adjust gain on the recording mic to move 20db to 0db (thus still giving 96db of range). As to how well a sound engineer *uses* that range....

[Pete Schumacher]

These days, the 90dB speaker is none too common.  Most offerings these days are in the mid to low 80s.  So for a full range speaker that meets the WAF form factor and can still do a reliable 30Hz in-room, you're looking at maybe 84dB/1W @ 1m.  This type of speaker is completely unsuited for use with SET amps if you sit more than a few feet away from the speaker.  Speakers like that need big power, or they simply start to sound harsh when turned up.  Then the listener will complain that said speaker sounds harsh when turned up, not realizing that he's clipping the amp 10% or more of the time.

If your room is somewhat "dead" you may need some extra amp headroom.  If you sit 20' away, obviously more headroom.

How much power you need is such an open ended question . . .

[steve f]

I checked and discovered that I rarely play at 95 Db peaks. As a matter of fact, I found that trying to play at 105 Db peaks was uncomfortable. Hooking up my old Knight scope proved I need very little power at all. With my efficient speakers, and allowing for a small amount of clipping, 1-3 WPC is plenty. I'm firmly in Roger's camp.

Steve

[Roger A. Modjeski]

I checked and discovered that I rarely play at 95 Db peaks. As a matter of fact, I found that trying to play at 105 Db peaks was uncomfortable. Hooking up my old Knight scope proved I need very little power at all. With my efficient speakers, and allowing for a small amount of clipping, 1-3 WPC is plenty. I'm firmly in Roger's camp.

Steve

Steve, Thanks and congrats for being the first person in this discussion to actually measure what is going into the speaker. What was the peak voltage you measured at your listening level and what was your "clip" voltage? Can you also tell us what your speaker sensitivity is, your listening distance and average SPL level?

What Steve did is the most realistic and accurate way to determine what you need. Who cares what your listening level is or speaker sensitivity. Now Steve knows what voltage he needs and that's that. He can also measure the current with the scope or meter or calculate it from the impedance curve of the speaker.

The smallest OTL I am making, and soon to offer to this group first, can supply 1 amp of current into any load from zero to 16 ohms and close to 1 amp all the way up to 32 ohms. As the load gets above 32 ohms the current falls off a little. From Steve's statement of 1 to 3 watts and assuming a 8 ohm speaker he can get 8 volts peak with the one amp of current.  That's well within his 3 watt upper limit. Unlike most other OTLs of the Futterman lineage this one has triode connected output tubes. Futterman's first amps were triode output but he never sold those commercially. Starting with the first commercial amp, the Harvard Electronics H-3a he was using pentodes in order to get high power. In addition to  the sonic advantage of triodes the amplifier is simpler and more reliable.

Although the power is interesting and is mentioned in almost every post, it's not what is most important. We don't listen to watts, we listen to  SPL. Thinking of light bulbs for instance do we buy watts or lumens? We really need to look at the lumens because the watts don't tell us how much light we are going to get.

All amplifiers act differently when driving a speaker and one should know a few things. For instance many Electrostatics go down to 1 ohm at high frequencies and need a fair amount of voltage at the same time. The Beveridge ESL amplifier produced 1500 Volt-amps (kinda like watts) and I measured 1500 VA while playing trumpet music at full output which for the 2-SW speaker was about 100 dB. Roger Sanders makes an enormous solid state amp for his ESLs for the same reason. You need a big one for the Martin Logans too. You don't need a very big one for the QUADS or my ESL as these speakers are nominally 8 ohms with a 4 ohm minimum.

People who know QUADS and don't want to burn them up use a suitable amplifier. The original QUAD amp was 15 watts as I recall and generally wired for 16 ohms. When all is said and done what the QUAD 57 wants is about 30 volts peak which is about what and RM-10 puts out of its 8 ohm tap. The RM-10 has become very popular among QUAD owners for that reason as it clips at just the right voltage to get the 57s up to full level and protect them at the same time without a clamp. The QUAD 63s want about 50 volts peak. So if i put my scope on the speaker terminals and get those peak numbers there is nothing more to safely get. In fact the 63 comes with a voltage clamp built in and many have added a clamp to the 57s to protect them. What this clamp does is essentially short out the amplifier above those voltages. A good analogy would be if you were driving along the freeway at 65 MPH and you wanted to drive at 55 you could leave the gas where it was and just use the brakes to hold you at 55. Now that would be a very bad thing for your engine, transmission and brakes. But essentialy that is what you do when you overdrive a QUAD. Once you reach the clamp voltage any extra voltage just heats up the clamp and your amplifier. Of course cone speakers can do this too by running out of excursion or with dynamic compression effects which DYNAUDIO has often pointed out in its advertising. So if you are a person who really likes high levels and headroom the speaker is just as important as the amplifier, certainly not enough is said about this. Just because your speaker says it can handle 100 or 300 watts that doesn't mean it is going to accurately reproduce it. Very few companies and no reviews I have seen show the effects of dynamic compression... Kudos to DYNAUDIO for doing so.. Of course they did it because they could.
 
I have to disagree with the posting that most speakers today are in the low 80 SPL for 1 watt. In my reading I see most conventional speakers in the high 80's. I think speaker sensitivity has improved over the years as driver manufacturers have learned how to make light cones sound good. I thank them for that because I can make a better sounding small amplifier than a large one and it costs less to own and operate. I note that most Wilson speakers are over 90 dB and his big ones are 93.5. Isn't it interesting that one of the biggest, heaviest most expensive speakers in the world is that efficient with a minimum recommended amplifier power of just 7 watts. Is Wilson trying to tell us something about small amplifiers sounding good?

I think we have come a long way from what impressed me in 1973 when I met my partner in what became Audio Art. He had put together his dream system. It was a pair of the big Magnepan Tympani 3's, thats 4 bass panels in the center and two midrange-tweeters on each side, 8 panels total. We drove it with a Phase Linear 700 on the bottom, Ampzilla (built from kit) on the mid and one of my early 100 watt transistor amps on the top. The crossover was a tube unit supplied by Audio Research (who was the sole supplier of the Tympani speakers). It was quite something for its time.

[JerryLove]

Steve, Thanks and congrats for being the first person in this discussion to actually measure what is going into the speaker. What was the peak voltage you measured at your listening level and what was your "clip" voltage? Can you also tell us what your speaker sensitivity is, your listening distance and average SPL level?

In fairness: he's the only person in this discussion with physical access to the speakers.

What Steve did is the most realistic and accurate way to determine what you need. Who cares what your listening level is or speaker sensitivity. Now Steve knows what voltage he needs and that's that. He can also measure the current with the scope or meter or calculate it from the impedance curve of the speaker.

I think that's overly dismissive of SPL measurements. If his source (for example) didn't include much LF, when he does include that it may up his power consumption. Also: it's terribly useful information when *selecting* speakers to know what your desired listening level will be.

Steve definitely did the right thing: and now he knows what his real needs are. Go Steve!

[rbbert]

It's all fine and good to spec SPL for a given voltage.  But music isn't a sine wave.

If you like the average level of your music at 90dB, a good recording is going to have as much as a 10:1 crest factor and more, which means instantaneous peaks to 100dB (string plucks, drum smacks, orchestral crescendos).

You always need more power than you think in order to completely avoid clipping of the amplifier during actual musical program material.

IME, well recorded, slightly compressed music has peaks that run 15-18 dB above average rms.  Many modern pop/rock recordings may have peaks only 4-6 db above average.

[Russell Dawkins]

From my experience, I am skeptical of any kind of power or level measurement being taken as gospel.

Years ago I was recording solo piano and heard clipping which indicated true levels at least 5-7dB over zero if in the digital domain and yet all meters in the chain (and it was a top quality chain) showed at least 10dB of headroom. By all meters, I refer to those on the mic preamp, the A/D converter (a $5000 converter) and the recording device, a Tascam DA30 DAT recorder. I had to back off on the gains until I was showing around 15dB of headroom at all stages before I got rid of the clipping distortion.

Looking into it later, I learned that level representations were almost always done from measurements of voltage in the analog domain, even with digital equipment.

Practically, it would appear that many a level meter under-represents the true peak levels of transients by as much as 15-20dB.

Since this event I have always been very wary of levels when recording piano since the waveform is very percussive, yet the sound is so smooth that any distortion is clearly heard.

[steve f]

I hope I got this right.

I have a pair of 8 Ohm 95 Db speakers. They aren't quite full range units, but are about 6 Db down at 35Hz. Our usual listening position is about 10 feet away.  Since OTL amps are becoming  a hot topic here, I used mine for the test. Now please accept that I was pretty casual about this. Heck, I was amazed that the Knight scope worked. It was a $20 garage sale purchase.

I found out that  2 volts was a good listening level, going up to 3 for my max. Sorry but I didn't pay any particular attention to where clipping got noticeable. The amp probably maxes out at about 1.25-1.5 WPC at 8 Ohms.

I expected similar results, because although I have several big solid state amps around, two with meters; the meters really never moved much.  I believe small amps do tend to sound better, and I admit that in the past I bought much more amp than I needed.

Additional info. I use a subwoofer, but turned it off for this experiment. My speakers operate close to the front wall in half PI space. I just grabbed an old Dave Grusin jazz CD for listening.


From what I observed, Roger's smallest OTL should be perfect for my system.

Steve

[JLM]

I have to disagree with the posting that most speakers today are in the low 80 SPL for 1 watt. In my reading I see most conventional speakers in the high 80's. I think speaker sensitivity has improved over the years as driver manufacturers have learned how to make light cones sound good. I thank them for that because I can make a better sounding small amplifier than a large one and it costs less to own and operate. I note that most Wilson speakers are over 90 dB and his big ones are 93.5. Isn't it interesting that one of the biggest, heaviest most expensive speakers in the world is that efficient with a minimum recommended amplifier power of just 7 watts. Is Wilson trying to tell us something about small amplifiers sounding good?

Don't believe everything you read.  Not uncommon for manufacturer's to boost the efficiency ratings by 3 dB or more (just one of those bench versus real world or more often simply a marketing ploy, just like amplifier vendors).  And take minimum amp recommendations with another grain of salt.  They might work playing at very low spls in a closet with poor resolution (amp lacks a commanding grip to have control of the drivers) all the while trying to avoid clipping. 

Where I've witnessed speakers improve in the last 40 years is a trend to provide deeper and tighter sound from fewer drivers in smaller cabinets (tighter coming from adopting European tastes).  But my experience is that highly efficient designs (sans the uber expensive models) still almost universally lack of deep bass and/or exhibit colorations (honky/forward midrange, tone is just plain off).

[JerryLove]

Don't believe everything you read.  Not uncommon for manufacturer's to boost the efficiency ratings by 3 dB or more (just one of those bench versus real world or more often simply a marketing ploy,

So I thought I might do an unscientific test. So I pulled up Stereophile (who does SPL measurements in their reviews) and thought I might just look at the first 5 speakers.

#1: My estimate of the Alexia's voltage sensitivity was slightly higher than the specified 90dB/2.83V/m, at 91.3dB(B)/2.83V/m.

#2 [Swans Speaker Systems Baton] B-weighted sensitivity weighed in at a moderate 85dB/2.83V/m. [manufacturer claim is 90db @1w]

#3 The T2 is specified as having an anechoic voltage sensitivity of 88dB; my estimate was slightly lower, at 87.3dB(B), almost within the margin of error.

#4 The Sonus Faber's voltage sensitivity is specified as 89dB/2.83V/m; my estimate was a little lower than that, at 87.7dB(B)/2.83V/m

#5 My estimate of the Wharfedale's voltage sensitivity was a fairly high 89.7dB(B)/2.83V/m, which is within experimental error of the specified 90dB.

So in 5 speakers (the first 5 on stereophile's webpage: no cherry-picking done)
0 were below 85db
1 had a spec that was more than 1db (rounding measurements) from measurement.

That's certainly not a statistically good group (since it's 5 and they are not completely random, though not chosen for efficiency nor honesty); but what data do you have that's better?

just like amplifier vendors).  And take minimum amp recommendations with another grain of salt.  They might work playing at very low spls in a closet with poor resolution (amp lacks a commanding grip to have control of the drivers) all the while trying to avoid clipping.

I don't believe this claim is any more accurate than your other.

[rbbert]

If you don't ever listen to music with 105 dB peaks, you either listen very quietly or to very compressed music.  Remember, most SPL meters read averages, and as noted earlier in this thread, even "peak reading" ones are probably reading somewhere below actual peaks on dynamic recordings.

[mcgsxr]

I have had high efficiency low power systems in the past (97db open baffle Visatons with 3-5wpc SEP Magnavox amp) and currently run a more conventional setup (B&K 140wpc and Paradigm 20's around 86.5db as measured by Stereophile) but don't know how much power I am actually using.

I tend to listen with peaks in the 85-90 range according to a free app on my daughter's iPad mini.

I'd assume most of the time I am using under 25w, but when there are big dynamic swings and I have it louder, it is nice to hear it all pull through fine.  I use the setup for HT too, and will listen to movies peaking in the mid to high 90's at times.

Roughly 8-9 foot listening distance, in a room 24x20 or so.

[rbbert]

having someone talk to you in a fairly normal speaking voice gives peaks of about 85db if you are about 3 ft from the person.