Well it's not as if any harm has been done
Although it would be interesting to see what phivates results would have been like if he'd listened to the symphony via a CD player + Pano's test the as the last movement is certainly quite dynamic.Yeah, not many folks do the workback from speaker efficiency to required power---they of course vary from person to person but typical listening levels are generally around 10mW RMS. As such, most speaker and amp design is really all about accommodating the times when you've got it cranked. Which is fine if cranking it is important, but increases cost and complexity and generally decreases performance in non-cranked situations.
In my case I did this level measuring exercise before building amps, with the result my dipoles are triamped with LME49600s. The dipole corrections and phase linearization mean my voltage envelopes don't translate easily to other systems but my normal listening's around 100mV on the tweeters, 250mV on the mids, and 200mV or so on subs. The LME49600s' current limiting means clipping happens at 1V on the tweeters, 2V on the mids, and generally around 2V on the subs.
+1
all the data is there and still you don't realize how little you need until you actually measure.
to tell a little story, I had this defective D/A converter that had a weird background noise (I returned it to the manufacturer who repaired it so I never got to find what the problem was). the noise was barely detectable with music even at low levels but on quiet passages it was really annoying. I used my trusty old mic input on my laptop to see what was actually there (the noise was low frequency so no risk of aliasing). to my surprise I found that the peak-to-peak level was 6mV! I simply couldn't believe that I was easily detecting some 10uW (that is microwatt) by ear from 2 meters in a not perfectly silent room. if you do the math you get ~15dB at listening position from 83 dB W/1m. I can't find the absolute threshold of hearing for low frequencies (for some reason Wiki only has a plot for 1kHz up) but it would seem reasonable that 15dB is above the threshold of detection at least for part of the band of that noise. but still 10 uW sounds (pun unintended) like very little, doesn't it?
Last edited:
why do you find it so hard to believe that some people simply don't need 105dB at listening position?
Last edited:
I typically use 1uW as a lower figure of merit---for the average 90dBish efficient driver that tends to produce soft whisper levels (30dB SPL) at the listening position and hence approximates background noise levels in a relatively quiet listening environment. Use nW occasionally and fW came up over here.
How's that compare to the mic input's noise floor? Generally integrated audio is not especially quiet.
ISO 226 and Fletcher-Munson curves are here.
Wrong, I'm afraid - because we don't know the recorded level of that cord. Keep the volume there, play the test tone and measure it. Then we'll know.
On your system, no. On my system 8 watts would be enough to hit very hard.
Remember, I get ~85dB at the listening seat with 1V into my system. The necessary amp output voltage is something each person can determine for himself, once the measurement is done.
Yes, the test is simple. But don't forget, I've asked for the loudest level you ever use. Most of the time you won't be anywhere near that voltage. And I would hope that even with whatever results you find, you get an amp that has a little more headroom, it can't hurt. Most folks are smart enough to figure that out. The difference between a 20WPC amp and 30WPC is less than 3 volts.
Ah, OK. I didn't get that part. Thanks.
With an analog source and a scope, you can still do this test. Just set your level to very loud, as loud as you ever go, then put the scope on the speaker terminals. Look for the peaks.
There you have it! If you divide that peak voltage by 4 (12dB) you'll end up at the same place as those using the digital test tone. You've measured your peak voltage directly, with the test tone we have to multiply by 4 to get the peak value.
femtowatt, now that is one unusual word
you wouldn't believe it and neither did I. I had tested the mic inputs on a few laptops before but hadn't found a half usable one until this one on the Pavilion g6 series. noise floor is at about -100 dB. others parameters are decent too.
I could be mistaken but as far as I remember absolute threshold of hearing and equal loudness curves are two different things. please correct me if I'm wrong.
Last edited:
What a game!
Anyway, lets consider a few things.
With a typical AC voltmeter or RMS Voltmeter, one can’t measure anything but a continuously signal accurately and with a typical DMM, the reading is bouncing all over which reflects the constantly changing signal which is a signal like music or noise.
Meters like a Simpson 260 are more useful than most simple DMM’s for anything that changes level.
A meter like an HP-400 has enough bandwidth to measure audio signals (where as many utility meters don’t) but even it has a limited dynamic linearity as does the Digital HP3456.
In other words, if you measure music with these, you will get a reading to be sure but it can be anywhere from meaningless or must be interpreted at best.
The same thing applies for a sound level meter unless it specifically has a ‘peak hold” function like a B&K 2204 what it reads is an integrated level over some time period.
The same is true for all analogue VU meters and most digital level indicators. NONE of these indicate instantaneous peak levels and very often amplifier clip lights have a significant time constant before turning on.
You can assume that any level indicator is not showing you the peak levels unless it has been intended to do so such as the free Orban recording style meters I linked to several times here.
The bottom line is, you cannot tell anything about the peaks without examining them with software or a the analogue level with an oscilloscope.
For example, if one wants to tell the instantaneous maximum pressure of an “event”, one analyzes the voltage /time record (what an oscilloscope displays) and considers the mic sensitivity and so on.
Lets talk about hearing too.
Are we concerned with loudness or reproducing the actual signal?
What seems to us like “loud” is not always the same as what instrumentation reads, in fact they are different enough to make things weird.
Loudness is partly related to time.
We use the impulse response of a speaker to show the time view of it’s magnitude and phase response, in fact they are interchangeable changeable views of the same thing via FFT and inverse FFT. A perfect speaker has a perfect impulse response, if you feed a perfect speaker that impulse wave shape as a signal , record the mic signal and do an fft, you have it’s response. For a brief time, they measured speakers and rooms this way but all the energy is contained in a short period.
To the ear, this signal can be played at clip and it does not sound loud, just a tick or pop. While it requires all of the headroom your system has and doesn’t sound like much at all, the reason is the time is short.
This kind of signal is the most difficult and dynamic and makes up sound like Tick’s, pops, snaps, balloon pops and firecrackers etc.
Now, I know a couple of you guys have been fooling with DSP so one signal that would be fun to generate would be a 0dB broadband impulse with realistic band limits say 20-20KHz. This would be a great example of large transient that isn’t loud.
Obviously a steady sound level takes care of the time issue but now we find subjective loudness is also strongly related to both the frequency range and bandwidth of the signal.
For example, to be detectable against a silent background, a signal at 20 Hz has to be about 10,000,000 times more powerful, than for a tone to be detectable at 3KHz.
So with DSP one can manipulate dynamics, what would be illustrative of the next part is if someone can generate some pink noise in several flavors of peak to average ratio. I suggest perhaps several other test signals could be made which will be more representative of the range of possible signals a speaker might be asked to reproduce and yet somewhat measurable at least with analogue RMS meter.
The AES standard type has a +6dB peak to average meaning that to have say 5V RMS with this signal, requires an amplifier that can put out twice the RMS voltage to accommodate the peaks.
What we hear as loudness is more or less this RMS or average level too, it is this (raising the average signal level with compression) that was manipulated to make TV commercials louder while not exceeding 0dB and now is called the CD loudness war. .
The noise signals generated by most RF test gear as well as Smaart and others has a 10dB peak to average ratio and is much more like compressed music than the 6dB variety. Here to deliver 5 VRMS to a load requires an amplifier that can deliver 16.6VRMS. While the peaks are larger, they are offset by deeper low spots, it is still the average level that more or less is “loudness”
So with DSP one of the curious here might be able to take normal pink noise and expand it’s dynamic range and lower the average level so that 0dB digital is never exceeded.
I would suggest then that pink noise tracks with peak to average ratio's at the normal 6dB, 10dB also 20dB, 30db and 40dB to emulate the most dynamic digital recordings (the medium can capture up to about 96dB dynamic range).
Oh, dB add quickly, to put out a paltry 5VRMS with that last pink noise signal with the 40dB peak to average ratio, takes an amplifier that can put out 500VRMS
It looks like Cool edit could do this (dynamic manipulation) but I have not tried it and don’t have time tonight but maybe someone here could.
In each case, the perceived loudness of each pink noise signal should correspond more or less to the average level and because the spectrum is the same for each, that should help make them more comparable.
In each case, the peak level is the same while the subjective loudness (average level) as well as the power limited maximum loudness changes (downward) by 34dB .
What is used for playback material can be anything from more or less continuous tones at low frequencies to the most difficult of all transients.
Only an examination of the dynamic signal instant by instant reveals what is happening instant by instant. For that, you need to look in the digital domain with software (if a digital signal) or in real life with something like an oscilloscope with sufficient bandwidth (250KHZ minimum) to show fast audio things.
While those short transient things don’t sound loud, they are part of what can make things sound real and well after they are lost dynamically, audible clues begin to tell you the system is straining.
Best,
Got to run
Tom Danley
Danley Sound Labs
Anyway, lets consider a few things.
With a typical AC voltmeter or RMS Voltmeter, one can’t measure anything but a continuously signal accurately and with a typical DMM, the reading is bouncing all over which reflects the constantly changing signal which is a signal like music or noise.
Meters like a Simpson 260 are more useful than most simple DMM’s for anything that changes level.
A meter like an HP-400 has enough bandwidth to measure audio signals (where as many utility meters don’t) but even it has a limited dynamic linearity as does the Digital HP3456.
In other words, if you measure music with these, you will get a reading to be sure but it can be anywhere from meaningless or must be interpreted at best.
The same thing applies for a sound level meter unless it specifically has a ‘peak hold” function like a B&K 2204 what it reads is an integrated level over some time period.
The same is true for all analogue VU meters and most digital level indicators. NONE of these indicate instantaneous peak levels and very often amplifier clip lights have a significant time constant before turning on.
You can assume that any level indicator is not showing you the peak levels unless it has been intended to do so such as the free Orban recording style meters I linked to several times here.
The bottom line is, you cannot tell anything about the peaks without examining them with software or a the analogue level with an oscilloscope.
For example, if one wants to tell the instantaneous maximum pressure of an “event”, one analyzes the voltage /time record (what an oscilloscope displays) and considers the mic sensitivity and so on.
Lets talk about hearing too.
Are we concerned with loudness or reproducing the actual signal?
What seems to us like “loud” is not always the same as what instrumentation reads, in fact they are different enough to make things weird.
Loudness is partly related to time.
We use the impulse response of a speaker to show the time view of it’s magnitude and phase response, in fact they are interchangeable changeable views of the same thing via FFT and inverse FFT. A perfect speaker has a perfect impulse response, if you feed a perfect speaker that impulse wave shape as a signal , record the mic signal and do an fft, you have it’s response. For a brief time, they measured speakers and rooms this way but all the energy is contained in a short period.
To the ear, this signal can be played at clip and it does not sound loud, just a tick or pop. While it requires all of the headroom your system has and doesn’t sound like much at all, the reason is the time is short.
This kind of signal is the most difficult and dynamic and makes up sound like Tick’s, pops, snaps, balloon pops and firecrackers etc.
Now, I know a couple of you guys have been fooling with DSP so one signal that would be fun to generate would be a 0dB broadband impulse with realistic band limits say 20-20KHz. This would be a great example of large transient that isn’t loud.
Obviously a steady sound level takes care of the time issue but now we find subjective loudness is also strongly related to both the frequency range and bandwidth of the signal.
For example, to be detectable against a silent background, a signal at 20 Hz has to be about 10,000,000 times more powerful, than for a tone to be detectable at 3KHz.
So with DSP one can manipulate dynamics, what would be illustrative of the next part is if someone can generate some pink noise in several flavors of peak to average ratio. I suggest perhaps several other test signals could be made which will be more representative of the range of possible signals a speaker might be asked to reproduce and yet somewhat measurable at least with analogue RMS meter.
The AES standard type has a +6dB peak to average meaning that to have say 5V RMS with this signal, requires an amplifier that can put out twice the RMS voltage to accommodate the peaks.
What we hear as loudness is more or less this RMS or average level too, it is this (raising the average signal level with compression) that was manipulated to make TV commercials louder while not exceeding 0dB and now is called the CD loudness war. .
The noise signals generated by most RF test gear as well as Smaart and others has a 10dB peak to average ratio and is much more like compressed music than the 6dB variety. Here to deliver 5 VRMS to a load requires an amplifier that can deliver 16.6VRMS. While the peaks are larger, they are offset by deeper low spots, it is still the average level that more or less is “loudness”
So with DSP one of the curious here might be able to take normal pink noise and expand it’s dynamic range and lower the average level so that 0dB digital is never exceeded.
I would suggest then that pink noise tracks with peak to average ratio's at the normal 6dB, 10dB also 20dB, 30db and 40dB to emulate the most dynamic digital recordings (the medium can capture up to about 96dB dynamic range).
Oh, dB add quickly, to put out a paltry 5VRMS with that last pink noise signal with the 40dB peak to average ratio, takes an amplifier that can put out 500VRMS
It looks like Cool edit could do this (dynamic manipulation) but I have not tried it and don’t have time tonight but maybe someone here could.
In each case, the perceived loudness of each pink noise signal should correspond more or less to the average level and because the spectrum is the same for each, that should help make them more comparable.
In each case, the peak level is the same while the subjective loudness (average level) as well as the power limited maximum loudness changes (downward) by 34dB .
What is used for playback material can be anything from more or less continuous tones at low frequencies to the most difficult of all transients.
Only an examination of the dynamic signal instant by instant reveals what is happening instant by instant. For that, you need to look in the digital domain with software (if a digital signal) or in real life with something like an oscilloscope with sufficient bandwidth (250KHZ minimum) to show fast audio things.
While those short transient things don’t sound loud, they are part of what can make things sound real and well after they are lost dynamically, audible clues begin to tell you the system is straining.
Best,
Got to run
Tom Danley
Danley Sound Labs
The 0dB equal loudness curve is the threshold of hearing curve. Give or take the +-10dB or so variation among normal listeners, the loss of high frequency sensitivity with aging, and the long term consequences of listening to one's stero at 105dB.
Actually this isn't quite correct.
If you set the volume to very loud with the analogue source and scope the outputs of your amplifier, then you will see the magnitude of the dynamic peaks. This will tell you the required voltage swing your amplifier needs to be able to reproduce those peaks at your given volume setting.
Of course if the amplifier is clipping already then you're stuck.
Lets say the amplifier isn't clipping though and you divide this figure by 4. Sure you're going to get a voltage figure that represents 12dB less output, but this isn't anywhere close to being the same thing as your -12dB sine wave.
In the case of the analogue source the peaks you will have encountered would have been entirely without reference. This is the same as when you're playing back music with a digital source and setting your volume control. We have no idea what the peak values are within the music but this is unimportant.
In the digital test once we've set the volume we then introduce a sine wave that has been recorded at -12dB as referenced from digital zero. This brings about a controlled element that we can now use as a reference for determining the maximum voltage swing that your system will ever be called upon to reproduce using a digital source. If we didn't know it's 12dB below the max it would be completely meaningless.
With the analogue source we don't have a reference point, those peaks we measured before, what did they represent? Nothing, just an arbitrary voltage output by the analogue source on a voltage peak. Who knows what this actually represents. You may have set the volume very loud but the next analogue piece of music you play could have peak levels significantly above the test you did before, we don't know, so if you did use the same volume setting you're amplifier could very well now be clipping.
If however you were trying to get a rough idea about the maximum voltage swing you typically use with your system, when using an analogue source at your loudest volume ever, then scoping the outputs would be decent. But as the levels are completely without reference then it could only be used as a very rough guide, rather then something absolute as with the digital system.
If though, lets say you scoped 10 volts peak with an analogue source (and I will say here for clarity as to avoid confusion, we are assuming this 10v peak represents what would be a peak as produced by a 0dBfs signal), first we need to convert it into an rms figure as the sine wave test is done to an rms value too. After that we can divide it by 4 so that it represents the same level as per the -12dB test.
I must confess Pano, I thought you were asking people to multiply the test tone voltage by 4 so the poll voltage actually represents the maximum rms voltage for their system, rather then the -12dB value. This I can only imagine might have added to the confusion about where people were getting the 4x peak to average idea from.
that never occurred to me but you're right. looking at the curve it makes sense that I heard those 15 dB's. the components seemed to be above 150 Hz.The 0dB equal loudness curve is the threshold of hearing curve. Give or take the +-10dB or so variation among normal listeners, the loss of high frequency sensitivity with aging, and the long term consequences of listening to one's stero at 105dB.
judging by some of the replies you could be right, some of them seemed to be implying that the chosen -12dB signal is related to the peak to average ratio in music.
if I read correctly the poll values are the actual value read by the DMM, not that value x4, right? that's how I voted.
Well if you paid a bit of attention and the reason why Pano test is so simple and elegant is that he chose 2 low frequency test
one that most certanly work and one that shuld work with beter meters.
The rest is just a game of egos you like toplay.
The rest of your post state preaty much the obvious I do not have oshiloscope Free Virtinis software or fully licenced Arta and fuly licenced TRueRTA or an Audiophile 192 sound card or I have chosen not to use it
Thanks for the test Mooly.
That contrasts with your 5V RMS measurement when you set the level by ear, right?
From the two pink noise at 85dB tests so far we see that I hit the same volume setting and you actually set yours 2.6dB lower. I'm beginning to be a lot less worried about the "you won't set it loud enough" complaints.
All within a good margin of experimental error I would say.
The old "doubling of power and 3db increase" saying, and of that 3db increase not being very noticeable is often quoted (not least by me) and that seems to be the case here. A small turn of the VC would have brought that 3.7 up to 5 volts and probably not have sounded much louder.
When I did the first test a few days ago I said I really cranked the level up
All good stuff
I only read the first 100 or so replies, so I'll be replying directly to the OP, and not to any of the further discussion towards the end of the thread.
After listening to some moderately dynamic music (some mixed electronic and acoustic chillout music - less dynamics than classical music, but more dynamics than pop or club music), I set to the maximum level that I'm likely to listen to on my own. The -9dB Peak sine wave (220Hz, since 120Hz is right where my crossover is) measured about 3.5V, meaning that the maximum possible digital sine wave at this gain level is about 10V, or 12.5W into an 8Ohm resistor.
My speakers I am using are Zaph Mini Bookshelfs with a 4" MCM Woofer and a quoted sensitivity of 86dB (1m,2.83V). 10V input is about 11dB louder than 2.83V input, so the maximum level in my room is about 97dB (at 1m, but considering the short critical distance in my tiny room, pretty much everywhere else as well). 10V input also leaves the speakers below xmax throughout the full spectrum (after 1st order highpass at 120Hz is applied), which is nice to see.
My subwoofers have varying levels of gain applied though, maximal gain to one of the woofers is 4.5dB at 50Hz, which means peak voltage is 1.67 times higher than to the speakers. So that's 5.9V for the -9dB Peak sine wave, or about 16.7V at maximum digital level. That's about 35W into 8 Ohm. I selected the "2-5V" option rather than the "5-10V" option, since I figured my mains would be more interesting than my subwoofers.
Some info on the side: The speakers are in a tiny room, approximately 27 cubic meters, with about 11 square meters area. I could probably listen a lot louder outdoors, but in here that's already quite loud.
After listening to some moderately dynamic music (some mixed electronic and acoustic chillout music - less dynamics than classical music, but more dynamics than pop or club music), I set to the maximum level that I'm likely to listen to on my own. The -9dB Peak sine wave (220Hz, since 120Hz is right where my crossover is) measured about 3.5V, meaning that the maximum possible digital sine wave at this gain level is about 10V, or 12.5W into an 8Ohm resistor.
My speakers I am using are Zaph Mini Bookshelfs with a 4" MCM Woofer and a quoted sensitivity of 86dB (1m,2.83V). 10V input is about 11dB louder than 2.83V input, so the maximum level in my room is about 97dB (at 1m, but considering the short critical distance in my tiny room, pretty much everywhere else as well). 10V input also leaves the speakers below xmax throughout the full spectrum (after 1st order highpass at 120Hz is applied), which is nice to see.
My subwoofers have varying levels of gain applied though, maximal gain to one of the woofers is 4.5dB at 50Hz, which means peak voltage is 1.67 times higher than to the speakers. So that's 5.9V for the -9dB Peak sine wave, or about 16.7V at maximum digital level. That's about 35W into 8 Ohm. I selected the "2-5V" option rather than the "5-10V" option, since I figured my mains would be more interesting than my subwoofers.
Some info on the side: The speakers are in a tiny room, approximately 27 cubic meters, with about 11 square meters area. I could probably listen a lot louder outdoors, but in here that's already quite loud.
jwmbro: Thanks very much for your results.
View Poll Results: I measured the test tone at:
I chose that wording carefully. Also see post #2, the actual test procedure:
All fine and dandy, Tom, but it has little to do with the test here. We are NOT measuring any music signal, just a sine wave. I chose a sine wave for the reasons you cite above. Please have a look at the first two posts in this thread.
Why the confusion? Look right at the top of the page where is says
View Poll Results: I measured the test tone at:
I chose that wording carefully. Also see post #2, the actual test procedure:
I thought that was pretty clear, maybe it wasn't.