Pretty darn sure the huge peak(s) in the second shot were snaredrum-hits.
The second positive peak could be some ghosting-effect due to the
camera's shutter-time interacting with the fast changing picture.
The two shots were taken from ~20 'random' shots i took using the delayed shutter-option on my camera,
to avoid movement of the camera due to me operating it.
If you don't mind i took a different route trying to correlate with your test-tone.
My neighbours hate me already for obvious (noisy) reasons.
Since the scope was set at 2 ms/div when taking the pictures, i took a ~20ms sample of a snaredrum-hit in Cooledit,
and overlayed your test-tone.
The second positive peak could be some ghosting-effect due to the
camera's shutter-time interacting with the fast changing picture.
The two shots were taken from ~20 'random' shots i took using the delayed shutter-option on my camera,
to avoid movement of the camera due to me operating it.
If you don't mind i took a different route trying to correlate with your test-tone.
My neighbours hate me already for obvious (noisy) reasons.
Since the scope was set at 2 ms/div when taking the pictures, i took a ~20ms sample of a snaredrum-hit in Cooledit,
and overlayed your test-tone.
Attachments
"or...quieter.....tracks don't go anywhere near it".
I would rephrase that to say :
or quieter tracks rarely go anywhere near it, and just very occasionally the recording technician will use virtually the full Dynamic Range available in the recording medium to maximise the S/N ratio of that quieter track. It may be that he goes to -0.1dBfs just once in a 3minute music track.
I've posted this story before.
I attended a school event where the lunchtime entertainment was a class of "African Drumming"
I did not realise that the teacher was leading the beat by repeatedly blowing on a whistle. For 99.99% of the time that whistle beat was completely drowned out.
As I was leaving I turned to a colleague and said. I wish my stereo could reproduce that, but I know it can't.
I attended a school event where the lunchtime entertainment was a class of "African Drumming"
I did not realise that the teacher was leading the beat by repeatedly blowing on a whistle. For 99.99% of the time that whistle beat was completely drowned out.
As I was leaving I turned to a colleague and said. I wish my stereo could reproduce that, but I know it can't.
Kvholio - excellent! Thanks for the screenshot and overlay. Looks just like I would expect, music down near the test tone and the snare hit up at 0dB. I.E., about 4X the voltage. I think it was higher on your actual scope shot, but can't be sure. I'll look at it in more detail this weekend.
Andrew. Yes, I definitely have some recordings were the average level is low, like -35dB or lower, but with peaks at -1dB. Some are classical recordings, other are just high dynamic range - including a few made by SY. However, I don't play those -35dB sections loud enough to rock the house because they are supposed to be quiet. If I did play those sections at 85dB SPL, then those peaks would take a beast of an amp. But for the most part, those sections just aren't played super loud, so the peaks are still within reason, like 105dB SPL or less.
As a number of posters here have wisely pointed out, most of us know very well the normal setting of our volume controls. Mine has never gone past 12:00, which is where I set it for my test. So I know the maximum voltage output of my amp at my maximum setting (12:00). For the most listening it's going to be a lot less. It will never be higher, unless I turn it up past 12:00. And so far, I never have, not for any recording.
Andrew. Yes, I definitely have some recordings were the average level is low, like -35dB or lower, but with peaks at -1dB. Some are classical recordings, other are just high dynamic range - including a few made by SY. However, I don't play those -35dB sections loud enough to rock the house because they are supposed to be quiet. If I did play those sections at 85dB SPL, then those peaks would take a beast of an amp. But for the most part, those sections just aren't played super loud, so the peaks are still within reason, like 105dB SPL or less.
As a number of posters here have wisely pointed out, most of us know very well the normal setting of our volume controls. Mine has never gone past 12:00, which is where I set it for my test. So I know the maximum voltage output of my amp at my maximum setting (12:00). For the most listening it's going to be a lot less. It will never be higher, unless I turn it up past 12:00. And so far, I never have, not for any recording.
I finally got to the point to test mine.
Results:
120 hz test tone - 0,653V
220 hz test tone - 0,653V (surprise)
speeakers - Dali ASX 8000 (92 dB, 5 Ohms)
amp - B&O 125ASX2 board (I just noticed that I have a thing for danish products with ASX designation
)
source - Sonos ZonePlayer 90
I voted 2V or less
Jakub
Results:
120 hz test tone - 0,653V
220 hz test tone - 0,653V (surprise)
speeakers - Dali ASX 8000 (92 dB, 5 Ohms)
amp - B&O 125ASX2 board (I just noticed that I have a thing for danish products with ASX designation
)source - Sonos ZonePlayer 90
I voted 2V or less
Jakub
For fun, lets look at this issue from a different angle.
Lets pretend we used the goal of a home theater as the target we wish to hit. This may be a bit louder than a home stereo but at least there are numbers
http://www.dolby.com/uploadedFiles/...lish_PDFs/Professional/L.mn.0002.5.1guide.pdf
On section 3.3 it mentions each of the speakers need to produce 85dBC using an slm or RTA and the pink noise signal.
Depending which flavor of pink noise (there are two, one like the AES standard has a 6dB peak to average and like Smaart uses is a 10dB peak to average) one knows the peak levels were either +6 or +10dB greater than the average but not shown on the meters as they show level integrated over time (and would show up on an oscilloscope etc).
Lets say you had speakers that were 90dB 1w 1m sensitive and you say 12 feet or four meters away.
That distance due to the inverse square law places the level at the lp -12dB down from the level at one meter.
This means that to produce the required 85dB spl, it requires the speaker produce 97 dB which requires 5 Watts RMS BUT it takes an amplifier of either 20 or 50 Watts to reproduce the test signal used (the instantaneous peaks being either +6dB or +10 dB depending on the pink noise).
To be clear, Sound Level meters read a time integrated average even on a “fast” setting and what I am talking about is not loudness as we hear it but the dynamic content.
With cool edit, one can examine any part of a music track and read the statistics.
In looking through what I had in my computer, I found a few tracks that make good examples.
One of my daughters the delicate romp called Velvet revolver is a good place to start, track one. Examining the center chunk of the song, one finds a fairly hard core volume war recording.
The Average level is -8.3 below 0dB (the peak level) while the lowest level reached was-34.5dB
So, this track could be played a little louder than the 85dB IF one had the 50W amplifier but would be limited by instantaneous clipping to a level -2.3dB below 85dB with a 20 amplifier.
Now lets compare that to an older rock recording of my era, the Cream’s Crossroads.
The average level is -14.9dB from the peak level (0dB) while the quietest part is -55dB down from 0 dB. Obviously much more dynamic range even though it was an analogue recording from the 70’s
Now with a 20 Watt amplifier to avoid instantaneous clipping, the average level can be no higher than 76dB at the lp.
Lets take a good modern recording, Livingston Taylor’s version of Isn’t she lovely.
Here the average level is -19.5dB down from the peak level (0dB) and so with the same 20W amplifier, the average level could be no higher than 71.5dB spl.
Take an extremely dynamic commercial recording like ozone percussion group’s jazz variations.
It’s average level is -21dB down from the peak (and has 53 and 49 clipped samples).
The maximum average level for this is about 70dB or about -15 dB down from reference..
If you down load the Harley recording from our web site at work, one finds the average level is -27dB from the peak at 0 dB and the lowest level is -89dB
The dynamic requirements of this mean the 20 Watt amp the maximum level before clipping is 64dB, a whopping -21dB quieter than with the low dynamic range noise test signal
An extreme case to be sure but the fireworks recording also on the web site may be the hardest one of all to produce without clipping. That is very difficult both because of the low frequency extension and no compression.
The average level is -41dB down from 0dB and so with the 20W amp, the maximum average level is 50dB SPL max while the minimum level is about -92dB.
Now, understand when I say average level, that corresponds to how loud it seems to you, what it would read on a sound level meter.
On the other hand, the instantaneous peaks are part of the music or signal and while you can’t hear them when they are cutoff (like longer term clipping), you sure can hear the difference if you care able to compare to an unclipped version and A vs B.
The clipped version sounding less dynamic, not wrong or bad but compared to ‘without” not as dynamic.
The Harley and Fireworks recordings are at the bottom of this page. T
ry them both on headphones first, these are different and not using anything you would recognize to do it.
Danley | Technical Downloads
Hopefully this makes the issue of how much” you need a bit clearer or at least clear that it depends what your trying to do and what the program material is. Each time you increase the peak to average by 10 dB, it means the peaks need to be 10X larger to keep the same apparent loudness.
The ONLY way to know for sure your amp isn't clipping on the instantaneous peaks is to check with an oscilloscope.
Best,
Tom Danley
Danley Sound Labs
Lets pretend we used the goal of a home theater as the target we wish to hit. This may be a bit louder than a home stereo but at least there are numbers
http://www.dolby.com/uploadedFiles/...lish_PDFs/Professional/L.mn.0002.5.1guide.pdf
On section 3.3 it mentions each of the speakers need to produce 85dBC using an slm or RTA and the pink noise signal.
Depending which flavor of pink noise (there are two, one like the AES standard has a 6dB peak to average and like Smaart uses is a 10dB peak to average) one knows the peak levels were either +6 or +10dB greater than the average but not shown on the meters as they show level integrated over time (and would show up on an oscilloscope etc).
Lets say you had speakers that were 90dB 1w 1m sensitive and you say 12 feet or four meters away.
That distance due to the inverse square law places the level at the lp -12dB down from the level at one meter.
This means that to produce the required 85dB spl, it requires the speaker produce 97 dB which requires 5 Watts RMS BUT it takes an amplifier of either 20 or 50 Watts to reproduce the test signal used (the instantaneous peaks being either +6dB or +10 dB depending on the pink noise).
To be clear, Sound Level meters read a time integrated average even on a “fast” setting and what I am talking about is not loudness as we hear it but the dynamic content.
With cool edit, one can examine any part of a music track and read the statistics.
In looking through what I had in my computer, I found a few tracks that make good examples.
One of my daughters the delicate romp called Velvet revolver is a good place to start, track one. Examining the center chunk of the song, one finds a fairly hard core volume war recording.
The Average level is -8.3 below 0dB (the peak level) while the lowest level reached was-34.5dB
So, this track could be played a little louder than the 85dB IF one had the 50W amplifier but would be limited by instantaneous clipping to a level -2.3dB below 85dB with a 20 amplifier.
Now lets compare that to an older rock recording of my era, the Cream’s Crossroads.
The average level is -14.9dB from the peak level (0dB) while the quietest part is -55dB down from 0 dB. Obviously much more dynamic range even though it was an analogue recording from the 70’s
Now with a 20 Watt amplifier to avoid instantaneous clipping, the average level can be no higher than 76dB at the lp.
Lets take a good modern recording, Livingston Taylor’s version of Isn’t she lovely.
Here the average level is -19.5dB down from the peak level (0dB) and so with the same 20W amplifier, the average level could be no higher than 71.5dB spl.
Take an extremely dynamic commercial recording like ozone percussion group’s jazz variations.
It’s average level is -21dB down from the peak (and has 53 and 49 clipped samples).
The maximum average level for this is about 70dB or about -15 dB down from reference..
If you down load the Harley recording from our web site at work, one finds the average level is -27dB from the peak at 0 dB and the lowest level is -89dB
The dynamic requirements of this mean the 20 Watt amp the maximum level before clipping is 64dB, a whopping -21dB quieter than with the low dynamic range noise test signal
An extreme case to be sure but the fireworks recording also on the web site may be the hardest one of all to produce without clipping. That is very difficult both because of the low frequency extension and no compression.
The average level is -41dB down from 0dB and so with the 20W amp, the maximum average level is 50dB SPL max while the minimum level is about -92dB.
Now, understand when I say average level, that corresponds to how loud it seems to you, what it would read on a sound level meter.
On the other hand, the instantaneous peaks are part of the music or signal and while you can’t hear them when they are cutoff (like longer term clipping), you sure can hear the difference if you care able to compare to an unclipped version and A vs B.
The clipped version sounding less dynamic, not wrong or bad but compared to ‘without” not as dynamic.
The Harley and Fireworks recordings are at the bottom of this page. T
ry them both on headphones first, these are different and not using anything you would recognize to do it.
Danley | Technical Downloads
Hopefully this makes the issue of how much” you need a bit clearer or at least clear that it depends what your trying to do and what the program material is. Each time you increase the peak to average by 10 dB, it means the peaks need to be 10X larger to keep the same apparent loudness.
The ONLY way to know for sure your amp isn't clipping on the instantaneous peaks is to check with an oscilloscope.
Best,
Tom Danley
Danley Sound Labs
Hi Tom,
Those recordings are superb
I'll use them in the demo for the upcoming Ottawa DIY meet.
I do have a few points to bring up with what you wrote though:
1. The 85dBc is a Dolby spec, not necessarily what your average audiophile would consider a "good" listening level for music. I know that it's a spec, and that's probably better than nothing, but a big part of the equation is the level at which each individual is comfortable listening at.
2. The home theater room you describe would have to be 33 feet by 33 feet, or almost 1100 square feet. Probably not entirely realistic. I don't really think that being 4 meters away from each loudspeaker in an HT setup is a realistic distance. I would say 2-3 meters is proabably more like it.
3. I would also argue that the 6dB for doubling of distance is more indicative of a freefield environment, when your average home theater room would probably come closer to a diffuse field situation. That, along with the closer distance would drastically reduce the required output level from the loudspeaker.
4. The recordings you posted above are a lot of fun, but not really indicative of any sort of musical recording. This again alleviates the power requirements.
Overall, I get the impression that what you outlined is a good indication of a worst case scenario, but may not be entirely applicable to a more average situation.
Cheers,
Owen
Those recordings are superb
I'll use them in the demo for the upcoming Ottawa DIY meet.
I do have a few points to bring up with what you wrote though:
1. The 85dBc is a Dolby spec, not necessarily what your average audiophile would consider a "good" listening level for music. I know that it's a spec, and that's probably better than nothing, but a big part of the equation is the level at which each individual is comfortable listening at.
2. The home theater room you describe would have to be 33 feet by 33 feet, or almost 1100 square feet. Probably not entirely realistic. I don't really think that being 4 meters away from each loudspeaker in an HT setup is a realistic distance. I would say 2-3 meters is proabably more like it.
3. I would also argue that the 6dB for doubling of distance is more indicative of a freefield environment, when your average home theater room would probably come closer to a diffuse field situation. That, along with the closer distance would drastically reduce the required output level from the loudspeaker.
4. The recordings you posted above are a lot of fun, but not really indicative of any sort of musical recording. This again alleviates the power requirements.
Overall, I get the impression that what you outlined is a good indication of a worst case scenario, but may not be entirely applicable to a more average situation.
Cheers,
Owen
I did this test pretty much as you described it (used a different test frequency though) a short while after I got my new speakers.
result was about 8-9V (~20W). I'll have to redo it though to be certain but unfortunately I have otitis right now and that would push the results up by few volts I think damned winter, we had -22 celsius yesterday and my ears hate it.
IMO the test would need to be done with worst case music, with a large crest (peak-to-average) factor. ok, someone said that on the quiet passages you're not supposed to crank the volume but to be frank, I think everyone does it
also, the music should be normalized. take for instance an album that I like and play a lot, Peter Gabriel - Security. it has a headroom margin of "only" -3dB (quoted from digido's Honor Roll) and sounds really quited if played after a compressed recording. it forces me to turn up the volume. oh, and it sounds very dynamic too.
quite the opposite of a compressed and clipped recording of a metal band.
my setup description:
room 4x3x2.5m
Dynaudio Confidence 3 speakers, specified at 86dB 2.83 V/m but with Dynaudio you never know, they are known for finding "legal" ways of pumping up the numbers
amp is "downpowered" (+/-37V supply) UCD180 - it starts clipping at about 22V in a 4.7 ohm resistive load, all harmonics rise abruptly around that voltage
impedance is 4 point something, flat from resonance to 20kHz. I measured it.
so I have about 2x100W of clean power with 2x80W of headroom and that is using 83dB/W/m speakers. kinda puts the typical "you need 500W minimum for dynamic peaks" BS in perspective. where are those people with 80dB speakers in huge rooms?
result was about 8-9V (~20W). I'll have to redo it though to be certain but unfortunately I have otitis right now and that would push the results up by few volts I think
damned winter, we had -22 celsius yesterday and my ears hate it.IMO the test would need to be done with worst case music, with a large crest (peak-to-average) factor. ok, someone said that on the quiet passages you're not supposed to crank the volume but to be frank, I think everyone does it
also, the music should be normalized. take for instance an album that I like and play a lot, Peter Gabriel - Security. it has a headroom margin of "only" -3dB (quoted from digido's Honor Roll) and sounds really quited if played after a compressed recording. it forces me to turn up the volume. oh, and it sounds very dynamic too.
quite the opposite of a compressed and clipped recording of a metal band.
my setup description:
room 4x3x2.5m
Dynaudio Confidence 3 speakers, specified at 86dB 2.83 V/m but with Dynaudio you never know, they are known for finding "legal" ways of pumping up the numbers
amp is "downpowered" (+/-37V supply) UCD180 - it starts clipping at about 22V in a 4.7 ohm resistive load, all harmonics rise abruptly around that voltage
impedance is 4 point something, flat from resonance to 20kHz. I measured it.
so I have about 2x100W of clean power with 2x80W of headroom and that is using 83dB/W/m speakers. kinda puts the typical "you need 500W minimum for dynamic peaks" BS in perspective. where are those people with 80dB speakers in huge rooms?
Last edited:
No, it's not. You're thinking bottom up, not top down. This is a top down test. If I do this test properly and get 1.5V on the test signal, then I'll never have an output of more than 4x1.5V or 6V peak. I know that my 30WPC amp isn't clipping at 6V. If it were, it would not be a 30W amp.
Yes, I have checked it with an oscilloscope. but I don't have to.
As regarding this test, please read what OPC has just posted - he makes some very good points.
Hi all,
this pair of results seems to predict that 0.653^2 = 0.43W amplifier would be required to drive an 8ohms speaker.
Jak is using a nominal 5ohms speaker, we are told that applying a factor of 2 corrects the prediction to suit a 4ohms speaker.
That results in a 0.86W Power Amplifier recommendation that will never clip on music signals reproduced from that digital source.
Have I interpreted that correctly?
Hi OPC
I am glad you like the recordings, that detection system invention is something I should get moving again, being a background project it has been going slowly. Unfortunately the two channels here only capture the view between two speakers so with headphones one looses the 360 degree around image that the other channels captured, these being only the front view, something similar to your vision’s field of view. AS you might have guessed, reproducing “reality” is a keen interest of mine and this includes looking at the dynamics of things.
As I mentioned, 85dbC is louder than some folks listen to the hifi but at least it has a number as opposed to a nebulous “pretty loud”. Conversely, some people have parties or company over in which case levels can be considerably higher too.
Fwiw, my speakers are 14 feet from the couch but my room is nothing like 33 feet by 33 feet but if one sits 2 meters away, just add 6 dB to the level.
Yes, room reflections can raise the level relative to the inverse square law but the more directional the speakers are the larger the direct field is and the less room corruption is relative to the direct signal. Ideally, one has mostly direct sound at the LP at least so far as preserving the recorded image.
The recordings I mention were both commercial recordings as well as the two I made, the commercial ones do represent those (underlining the large variability in dynamic range between say Velvet Revolver and James Taylor’s brother Livingston.
I guess the issue here may be if no one looks to actually see what the program material contains or what sounds in everyday life are, one can only go by what it sounds like and you can’t hear this problem single ended (you need to compare with and without). Also many modern recordings are so compressed to get a higher average level that there is no requirement for significant peak power and so there is no problem playing them back with modest headroom.
Actually it was recording that got me interesting in the dynamics issue, in the old days, one was up against a limited dynamic range of open reel tape but the 24/96 format is wonderful even 16/44.1 is good. I would urge anyone interested in this stuff who has a measurement microphone to start making their own recordings using your computer. If one gets a usb interface such as an M-audio fast track pro, one can record in 24/96 format. If one gets an old copy of Cool edit or other similar programs, one can read the data and extract the information on peak and average levels from the data.
Best of all, as you made the recording AND unlike commercial recordings, you know what it sounded like first hand and can compare the reproduction based on that. What is so hard I don’t suggest it, is making a stereo image but with just one flat omni mic (measurement mic) you can capture sounds in your home that will make the hair on the back of your neck stand up (if you do that physiologically).
Keep in mind, or at least it appears to me that commercial audio is more or less steered in the direction of the mass market and lately with the emphasis being on portability not fidelity or realism.
The loudness war ties in perfectly with the decreasing capability of portable devices and compact data compression.
If you are jogging or riding in a car, the background noise level is very high so compression raises the quietest parts to more like the background noise and the average above that. With a 60dB background noise level, 90dB of dynamic range is impractical to preserve and the average level stays buried in noise.
We have been using generation loss recordings at work to keep track of our progress as well as our competition. This isn’t done with loudspeakers so far as I am aware but it has been useful for us. The speaker is on a tower with a low noise measurement mic at a meter. The speaker is driven with a music track and the direct and mic signals are recorded on a 24/96 recorder. The mic signal is played back through the speaker at about 90dB and the music track is re-recorded given a gen loss comparison on the recording system.
With each generation, the speakers warts, even dynamic compression become more and more exaggerated. Funny, some speakers sound funky on one recording (no generation loss, just capture) while good ones can go three generations before starting to sounding funky.
For those who design loudspeakers, it is Quite a humbling reality check compared to electronics.
We did an outdoor demo in December that sort of captured the effect of this, one of the attendees was walking around with what looked like an I phone taking a video. Later when I saw the video and heard the audio, I asked if he could post it to the face book page. Here is a speaker that is a 4 way high directivity system that has 22 drivers but it radiates as if it only had a single driver, there is no comb filtering as you move around the pattern, no phase shift going through crossover..
Unlike large arrays, these are hardly effected by a cross wind and maintain the same spectral balance even over large distances.
Try this with headphones, adjust the volume to scale at 1:20 or so when the operator walks up to the fellow next to the camera man and talks. Unfortunately, the fellows mic’s were squashed when he gets too close to the subwoofer.
Danley Sound Labs - YouTube
As these are to be used in large scale sound like Imax theaters and stadiums, most of the people were standing on a ridge at 150 yards, at about 2:30 the camera man pans out that direction. The sensitivity of the speaker in the air is about 110dB 1W 1M, to produce the peaks and it’s drive by a large lab Grupen amp.
Hi Pano
I guess we are talking about two different things or have different requirements so far as what can be reproduced, what our systems are used for. My last writing started with the maximum power and worked downwards based on what the data in the recordings contained.
Either one clips the peaks (thankfully that doesn’t sound bad, just less dynamic), listens at a lower average volume or has enough power not to clip the peaks. If one is interested, then it is worth looking carefully. If knowing what a given signal is doing at a given time is important, the only way to know is to examine it.
Best,
Tom Danley
I am glad you like the recordings, that detection system invention is something I should get moving again, being a background project it has been going slowly. Unfortunately the two channels here only capture the view between two speakers so with headphones one looses the 360 degree around image that the other channels captured, these being only the front view, something similar to your vision’s field of view. AS you might have guessed, reproducing “reality” is a keen interest of mine and this includes looking at the dynamics of things.
As I mentioned, 85dbC is louder than some folks listen to the hifi but at least it has a number as opposed to a nebulous “pretty loud”. Conversely, some people have parties or company over in which case levels can be considerably higher too.
Fwiw, my speakers are 14 feet from the couch but my room is nothing like 33 feet by 33 feet but if one sits 2 meters away, just add 6 dB to the level.
Yes, room reflections can raise the level relative to the inverse square law but the more directional the speakers are the larger the direct field is and the less room corruption is relative to the direct signal. Ideally, one has mostly direct sound at the LP at least so far as preserving the recorded image.
The recordings I mention were both commercial recordings as well as the two I made, the commercial ones do represent those (underlining the large variability in dynamic range between say Velvet Revolver and James Taylor’s brother Livingston.
I guess the issue here may be if no one looks to actually see what the program material contains or what sounds in everyday life are, one can only go by what it sounds like and you can’t hear this problem single ended (you need to compare with and without). Also many modern recordings are so compressed to get a higher average level that there is no requirement for significant peak power and so there is no problem playing them back with modest headroom.
Actually it was recording that got me interesting in the dynamics issue, in the old days, one was up against a limited dynamic range of open reel tape but the 24/96 format is wonderful even 16/44.1 is good. I would urge anyone interested in this stuff who has a measurement microphone to start making their own recordings using your computer. If one gets a usb interface such as an M-audio fast track pro, one can record in 24/96 format. If one gets an old copy of Cool edit or other similar programs, one can read the data and extract the information on peak and average levels from the data.
Best of all, as you made the recording AND unlike commercial recordings, you know what it sounded like first hand and can compare the reproduction based on that. What is so hard I don’t suggest it, is making a stereo image but with just one flat omni mic (measurement mic) you can capture sounds in your home that will make the hair on the back of your neck stand up (if you do that physiologically).
Keep in mind, or at least it appears to me that commercial audio is more or less steered in the direction of the mass market and lately with the emphasis being on portability not fidelity or realism.
The loudness war ties in perfectly with the decreasing capability of portable devices and compact data compression.
If you are jogging or riding in a car, the background noise level is very high so compression raises the quietest parts to more like the background noise and the average above that. With a 60dB background noise level, 90dB of dynamic range is impractical to preserve and the average level stays buried in noise.
We have been using generation loss recordings at work to keep track of our progress as well as our competition. This isn’t done with loudspeakers so far as I am aware but it has been useful for us. The speaker is on a tower with a low noise measurement mic at a meter. The speaker is driven with a music track and the direct and mic signals are recorded on a 24/96 recorder. The mic signal is played back through the speaker at about 90dB and the music track is re-recorded given a gen loss comparison on the recording system.
With each generation, the speakers warts, even dynamic compression become more and more exaggerated. Funny, some speakers sound funky on one recording (no generation loss, just capture) while good ones can go three generations before starting to sounding funky.
For those who design loudspeakers, it is Quite a humbling reality check compared to electronics.
We did an outdoor demo in December that sort of captured the effect of this, one of the attendees was walking around with what looked like an I phone taking a video. Later when I saw the video and heard the audio, I asked if he could post it to the face book page. Here is a speaker that is a 4 way high directivity system that has 22 drivers but it radiates as if it only had a single driver, there is no comb filtering as you move around the pattern, no phase shift going through crossover..
Unlike large arrays, these are hardly effected by a cross wind and maintain the same spectral balance even over large distances.
Try this with headphones, adjust the volume to scale at 1:20 or so when the operator walks up to the fellow next to the camera man and talks. Unfortunately, the fellows mic’s were squashed when he gets too close to the subwoofer.
Danley Sound Labs - YouTube
As these are to be used in large scale sound like Imax theaters and stadiums, most of the people were standing on a ridge at 150 yards, at about 2:30 the camera man pans out that direction. The sensitivity of the speaker in the air is about 110dB 1W 1M, to produce the peaks and it’s drive by a large lab Grupen amp.
Hi Pano
I guess we are talking about two different things or have different requirements so far as what can be reproduced, what our systems are used for. My last writing started with the maximum power and worked downwards based on what the data in the recordings contained.
Either one clips the peaks (thankfully that doesn’t sound bad, just less dynamic), listens at a lower average volume or has enough power not to clip the peaks. If one is interested, then it is worth looking carefully. If knowing what a given signal is doing at a given time is important, the only way to know is to examine it.
Best,
Tom Danley
[didn't see anyone who replied to this, so even though it's a few days old I will go ahead and comment. sorry if it's a repeat]
There is a general confusion about voltage and power here. This test tries to provide a general number for the maximum output voltage needed to reproduce the peaks in a piece of music. This can then be used to determine what power rating an amp would have to possess in order to be capable of providing this voltage.
What you are pointing out is that the actual power output from a signal with a specific peak voltage can be higher that that of a single sinusoid with the same peak voltage. The simplest example is comparing the power output of a sinusoid at a given frequency with that of a square wave at the same base frequency and amplitude. As you know the square wave can be looked at as a sum of many sinusoids and yes, it does require more power to reproduce. But it does not require any more voltage to reproduce it over the single sinusoid example.
(Note that a triangular wave (also a sum of sinusoids) requires less power that a sinusoid with the same base frequency and peak voltage.)
In general the advertised "peak" power rating of an amp is determined by it's maximum output voltage.
The ability of an amp to generate the necessary current and dissipate the heat required to drive a sustained square wave (or other signal) into a resistive load at its maximum output voltage is that subject of another discussion entirely.
-bill
There is a general confusion about voltage and power here. This test tries to provide a general number for the maximum output voltage needed to reproduce the peaks in a piece of music. This can then be used to determine what power rating an amp would have to possess in order to be capable of providing this voltage.
What you are pointing out is that the actual power output from a signal with a specific peak voltage can be higher that that of a single sinusoid with the same peak voltage. The simplest example is comparing the power output of a sinusoid at a given frequency with that of a square wave at the same base frequency and amplitude. As you know the square wave can be looked at as a sum of many sinusoids and yes, it does require more power to reproduce. But it does not require any more voltage to reproduce it over the single sinusoid example.
(Note that a triangular wave (also a sum of sinusoids) requires less power that a sinusoid with the same base frequency and peak voltage.)
In general the advertised "peak" power rating of an amp is determined by it's maximum output voltage.
The ability of an amp to generate the necessary current and dissipate the heat required to drive a sustained square wave (or other signal) into a resistive load at its maximum output voltage is that subject of another discussion entirely.
-bill
Last edited:
If you want something that will really make the room and your loudspeakers shake.
Launch of Space Shuttle Endeavour on STS-123 (24 Bit) : Joe Shambro : Free Download & Streaming : Internet Archive
Launch of Space Shuttle Endeavour on STS-123 (24 Bit) : Joe Shambro : Free Download & Streaming : Internet Archive
Hi 5th
Thanks for the link, THAT really is what it sounds like.
The popping is actually sonic booms generated by the supersonic exhaust, in person that VLF content makes your pant legs flap.
I had worked on payloads flown on STS-7 and STS51a in the 90’s and that brought back a powerful sonic memory.
In the old days a recording guy named Bob Katz brought a recording of a shuttle launch to our booth at a trade show in Orlando and play it a couple times but I never got a copy. I had been able to start a small speaker division within the company making a motor driven subwoofer called a Servodrive, we had I think 6 or 8 of the Contrabass units which were -3 at about 16Hz and I remember we made it snow ceiling tile dust playing it.
For this recording, the peak to average is about 20dB which is a nice uncompressed live recording, the REALLY hard part to reproduce live is spectrum and level, the greatest level appearing to be around 10Hz or so.
In reality very very few systems could do justice to this sonic event at anything like a real level (one can scale this up to real by turning up the volume before the launch until the background voices and sounds are in realistic scale (and be sure to turn it down before the launch).
Again, thanks for the link.
Best,
Tom Danley
Thanks for the link, THAT really is what it sounds like.
The popping is actually sonic booms generated by the supersonic exhaust, in person that VLF content makes your pant legs flap.
I had worked on payloads flown on STS-7 and STS51a in the 90’s and that brought back a powerful sonic memory.
In the old days a recording guy named Bob Katz brought a recording of a shuttle launch to our booth at a trade show in Orlando and play it a couple times but I never got a copy. I had been able to start a small speaker division within the company making a motor driven subwoofer called a Servodrive, we had I think 6 or 8 of the Contrabass units which were -3 at about 16Hz and I remember we made it snow ceiling tile dust playing it.
For this recording, the peak to average is about 20dB which is a nice uncompressed live recording, the REALLY hard part to reproduce live is spectrum and level, the greatest level appearing to be around 10Hz or so.
In reality very very few systems could do justice to this sonic event at anything like a real level (one can scale this up to real by turning up the volume before the launch until the background voices and sounds are in realistic scale (and be sure to turn it down before the launch).
Again, thanks for the link.
Best,
Tom Danley
Absolutely correct. And that's why I asked for voltage, not power. Power is a very emotional issue, it seems.
If an amp is honestly rated, then it will be able to provide its rated RMS power without significant distortion or clipping on a sine wave. Peak power will be twice as much, simply because sine wave peaks are 3dB higher in voltage than the RMS average. You can figure out what that voltage is by using Ohms law. V = (W x R)2 That means Voltage = the Square Root of Watts divided by Resistance. For speakers we use Impedance in place of Resistance.
Example:
-
Q: You have an amplifier rated at 25WPC (RMS) into 8 ohms. What's its peak voltage?
A: 25*8=200 200sqrt=14.14 The amplifier can supply 14.14 volts RMS into an 8 ohm load playing a sine wave. Since we know that the peak of a sine wave is 3dB higher than its RMS value, that means the amplifier will supply at least 19.94 volts peak.
In this test you measure a sine wave tone that has an RMS value that is %25 of the highest peak possible on a digital recording. Knowing that, you can use Ohms law again to figure out what power that is. And you can figure out the power rating an amplifier would need not to clip at that volume setting, or any lower setting.
It's really very simple. You measure at 25% of full scale (absolute max) and do a little math to see how that relates to power ratings.
Yes. My test is for finding the peak voltage you need for a given level setting. It is intended for people listening to music in their typical situation. I asked for the level to be set "as high as you ever go" so that we would not under estimate that voltage.
Of course the test could be used for other types of systems and goals, as long as your source has an easily determined maximum. Digital recordings do.
Yes. In contrast my test does not need as many assumptions or calculations because you actually set the maximum volume before the measurement. The test tone is simply 12dBFS (-9dB RMS) below that, for convenience.
I have tried setting SPL a reference pink noise, but never felt it related well to how I listen on my system. Because of that, I developed this simple and clever test.
Yes, I agree. And with the test method in this thread, that is exactly what we are doing. The test signal is a pure sine wave with a known amplitude and frequency chosen specifically for accurate measurement with a voltmeter.
The signal is easy to measure and has a precisely known relation to all other levels on a digital recording.
To know about some other signal, we might want to analyze it in a different way. Let's look at that shuttle lift-off recording.
Pretty close. According to both Goldwave and Ocenaudio, it's about -16dB if you don't count the almost silent part before lift-off. See the numbers below.Tom Danley said:
We can also see that the loudest part is pretty badly clipped on the left channel. (See analysis and waveform below.) With some simple and free software we can analyze the recording. Recording a shuttle launch is hard, NASA won't give you a sound check.
Attachments
The fun of the shuttle launch track, at least for me, is if you were imagining that you were there on the day. Then turning up the volume so that the idle chatter of the crowd were at a realistic level and then boom, now that's a dynamic range! The actual noise of the shuttle launching isn't particularly interesting to me, comparing it to the crowd is though.
As Tom said, I am sure if you wanted to faithfully reproduce the aural event then you'd end up half shaking the house half to bits.
As Tom said, I am sure if you wanted to faithfully reproduce the aural event then you'd end up half shaking the house half to bits.
I have posted this article as a separate thread a while back: Heavy Load: How Loudspeakers Torture Amplifiers | Stereophile.com
IMO it's a must read.
the author addresses the issue of power requirement from the loudspeaker's perspective, i.e. how one can derive power requirements from the loudspeaker impedance data.
as stated before, it's not about power into the speaker but rather about what goes on inside the amp's output devices. I haven't thought how that translates to switching amps, obviously things are different with those and the analysis is done for linear amps only.
try to drop "peak to average" view on things for a bit and look at it this way. the author of the article derives what he calls the EPDR (Equivalent Peak Dissipation Resistance), that is the resistive load which causes the same dissipation on the output transistors as a specific loudspeaker. opposed to what one might expect, maximum dissipation doesn't occur at minimum impedance magnitude.
for the B&W 802D the EPDR minimum is 1.5 ohms and it occurs at 50-60 Hz. run a frequency analysis on music material and you'll find that there's almost always a lot of content around those frequencies. so there you have it. it's basically like driving a 60 Hz sine into a 1.5 ohm resistor. the amp will have to be capable of delivering a large current so that it doesn't break, enter in protection mode or enter current limitation because of a weak supply.
the most notable minimum seems to be around the woofer resonance, more precisely around the frequency where the phase has a minimum. but there's no way to tell visually how "bad" it is.
what all this means is that while the actual power that is dissipated on the speaker might seem low, the amp still has to do a lot of work if the speaker forces it to. useless work if you ask me but that's a problem for the speaker designers.
IMO it's a must read.
the author addresses the issue of power requirement from the loudspeaker's perspective, i.e. how one can derive power requirements from the loudspeaker impedance data.
as stated before, it's not about power into the speaker but rather about what goes on inside the amp's output devices. I haven't thought how that translates to switching amps, obviously things are different with those and the analysis is done for linear amps only.
try to drop "peak to average" view on things for a bit and look at it this way. the author of the article derives what he calls the EPDR (Equivalent Peak Dissipation Resistance), that is the resistive load which causes the same dissipation on the output transistors as a specific loudspeaker. opposed to what one might expect, maximum dissipation doesn't occur at minimum impedance magnitude.
for the B&W 802D the EPDR minimum is 1.5 ohms and it occurs at 50-60 Hz. run a frequency analysis on music material and you'll find that there's almost always a lot of content around those frequencies. so there you have it. it's basically like driving a 60 Hz sine into a 1.5 ohm resistor. the amp will have to be capable of delivering a large current so that it doesn't break, enter in protection mode or enter current limitation because of a weak supply.
the most notable minimum seems to be around the woofer resonance, more precisely around the frequency where the phase has a minimum. but there's no way to tell visually how "bad" it is.
what all this means is that while the actual power that is dissipated on the speaker might seem low, the amp still has to do a lot of work if the speaker forces it to. useless work if you ask me but that's a problem for the speaker designers.
Last edited:
It always surprises how the hi-end circles continue to put out loudspeakers that are ridiculously hard to drive. Of course those in the high-end circles might like this because it gives them a very good reason to buy that expensive amplifier with oodles of power.
From a technical stand point though a loudspeaker that's difficult to drive will increase an amplifiers distortion and theoretically make it sound worse. Worst of all if your amplifier cannot supply the current it's protection circuitry will come in and that will sound bad too. In the case of the B&W they could have added extra components to the network to smooth out the impedance but you rarely see this done. KEF and Dynaudio regularly use networks that do this.
From a technical stand point though a loudspeaker that's difficult to drive will increase an amplifiers distortion and theoretically make it sound worse. Worst of all if your amplifier cannot supply the current it's protection circuitry will come in and that will sound bad too. In the case of the B&W they could have added extra components to the network to smooth out the impedance but you rarely see this done. KEF and Dynaudio regularly use networks that do this.