Thanks Mooly.
The Phillips CD calls that sine "0dB" because it is the loudest possible sine wave.
Right. My tone will be 9dB below the reference tone on your CD because the CD reference tone is actually at -3dB from peak. Why? because you can't record a sine wave at 0dB, it would be a flat line! If the peaks of the sine are at 0dB, then the RMS value is -3dB. See images below. Confusing? Kinda.
The Phillips CD calls that sine "0dB" because it is the loudest possible sine wave.
I noticed that on your screen shots. Why is that happening? Did you measure at the CD output, or at the speaker terminals? When I scoped, I saw almost all peaks at 4V, max. But one track I saw 1 or 2 peaks at 6V. Kick back from the speaker? I don't know.
Attachments
Sorry Andrew, that quote above means nothing in a digital recording. 0dBFS is as high as the signal can go. Everything else is below that.
However I do agree that if you use a pink noise signal with an average to peak ratio of 6dB, then your peaks will be twice the voltage of the average. That means 4X the power. I don't see any problem with that. If the pink noise had an average to peak ratio of 12dB, then peaks would be 4 time the average voltage, or 8X the power. Pretty much what I've been saying all along.
Why is my use of -12dB and -9dB at odds with your understanding? Don't you understand the difference between Full Scale and RMS values? Certainly you do. Did you see the waveforms I posted above?
However I do agree that if you use a pink noise signal with an average to peak ratio of 6dB, then your peaks will be twice the voltage of the average. That means 4X the power. I don't see any problem with that. If the pink noise had an average to peak ratio of 12dB, then peaks would be 4 time the average voltage, or 8X the power. Pretty much what I've been saying all along.
Why is my use of -12dB and -9dB at odds with your understanding? Don't you understand the difference between Full Scale and RMS values? Certainly you do. Did you see the waveforms I posted above?
OK, thinking about it a bit more, I can see where the confusion might come from. If a CD has a tone at "0dB" then one would suppose that the peaks are at 0dB full scale.
If you measure that tone with a volt meter, you would see 3dB less than peak value, because the meter reads the RMS value, not peak. A 0dB peak sine has an RMS value of -3dB.
My test tones have an RMS value of -12dBFS That means 12dB below full scale. That will put them 9dB (RMS) below a "0dB" sine wave.
I have tried to make that abundantly clear throughout the thread, but it looks like I failed. "What we have here, is failure to communicate."
I'm off to work and will be tied up all week on a show. Don't know if I'll have time to check in. Have fun without me, try not to tear up the place too much!
If you measure that tone with a volt meter, you would see 3dB less than peak value, because the meter reads the RMS value, not peak. A 0dB peak sine has an RMS value of -3dB.
My test tones have an RMS value of -12dBFS That means 12dB below full scale. That will put them 9dB (RMS) below a "0dB" sine wave.
I have tried to make that abundantly clear throughout the thread, but it looks like I failed. "What we have here, is failure to communicate."
I'm off to work and will be tied up all week on a show. Don't know if I'll have time to check in. Have fun without me, try not to tear up the place too much!
The scope was measuring line out from the player. I don't think anything in the 2-20KHz CD domain could provoke a measurable kickback. You need to reverse the direction of current flow really fast in inductive components to see that and I don't think anything off disc could do that.
I'll have think on all this
The great thing about a thread like this is that it really gets you thinking.
As I mentioned earlier, 16 bit = 65536 discrete steps. The Philips disc also say this
"Level Definition"
The reference level of 0db refers to a sinewave whose most positive and negative value corresponds to +32767 quantisation steps and whose negative value corresponds to -32767 quantisation steps
So... and this is something I have never considered... is that saying the sinewave is "biased
" to the midpoint so that it can swing equally in each direction. A kind of digital midpoint ? Two times 32767 is 65534 discrete steps. I am sure there is some sound technical reason for it not being 65536.I honestly don't know
I will try the white noise on another player and see although surely it can't be an effect of the player.
Enjoy the show
OK, I just followed directions and performed the test. AC voltage at the speaker terminals measured 3.4V.
Equipment is a 6W per channel PP 2A3 amp into Snell Type J/III speakers.
--
Sorry, I forgot some of the other info requested.
- The room is 12' x 24' with an 8' ceiling.
- The speakers are touted to have 91dB/1w/1m sensitivity. On the high side of "standard."
So 3.4^2 = 11.56
I need a 12 watt per channel amp? Hmmm...
--
Equipment is a 6W per channel PP 2A3 amp into Snell Type J/III speakers.
--
Sorry, I forgot some of the other info requested.
- The room is 12' x 24' with an 8' ceiling.
- The speakers are touted to have 91dB/1w/1m sensitivity. On the high side of "standard."
So 3.4^2 = 11.56
I need a 12 watt per channel amp? Hmmm...
--
Last edited:
Hi
I think what was confusing at least for me was the title suggesting this indicated how much power your speakers need (like with music).
What it tells you is how much Voltage it takes to be subjectively pretty loud with a mid bass sine wave.
Depending on the music and recording, it may take 4, 10, 100, 1000 or more times that much (power) to encompass what is in some recordings if played at the same subjective average loudness.
It is like Pano said, digital zero is a point you cannot / should not pass, that level is the loudest / largest Voltage sample that can be recorded on that medium.
For a moment, pretend loudspeakers are dynamically linear, if you put in twice as much power, it puts out the signal 3dB louder and so on (which they do up to between about 1/10 and 1/8 rated power when they begin to compress).
Lets say one had speakers that were 85dB 1w 1m and one sat in an absorptive room (to make it simple) and were 4 meters from each speaker. As this is “stereo” we will take a worst case and say you have a signal on one channel.
Lets say you’re comfortable loud level was only 80dB SPL in that sine wave test.
AS your 4 meters away, due to the inverse square law we know the level is +6dB greater at one meter and we know that at one meter, the speaker was 86dB or 1dB more than 1W or 1.26 Watts.
To get the same SPL reading on a sound level meter using broad band AES pink noise at couple octaves or more wide, it takes an amplifier than can put out 55 Watts (RMS) or twice the Voltage as the signal has a 6dB peak to average ratio.
Similarly with a recording that has a 30dB peak to average ratio, one finds the peaks require 1260 Watts while the average level for the same SPL is still only 1.26 Watts.
Now, +30dB over 80dB sounds like it would be loud, and it would be if it were also a continuous sine wave but if used to reproduce a snare drum or other percussive instrument, it will fall FAR short of the real thing.
It is the very difficulty in capturing and reproducing large dynamic ranges, particularly in a high noise background, that lead to compression and then the low bit MP-3’s so you can hear music when your jogging or with the car top down.
Thankfully one can’t hear instantaneous clipping like traditional “ugh turn it down” clipping, this only saps the dynamics. The only way around this is to have efficient speakers and power or sit very close, or choose your music to fit the system’s capability.
Ultimately, with so many variables in equipment and program material, the ONLY WAY to tell if you have enough Voltage or Power (either view) is to examine what is coming out of your power amplifier with an oscilloscope. If you have one and know how to use it, DO try this, examine the highest peaks and see if any are flat topped.
More often than I would have thought, one finds instantaneous clipping and not just from a power amplifier but any stage can clip, the arrangement of this is “optimizing gain structure”.
Also, it’s worth pointing out that in the recording and engineering areas what we see as dynamic range can be referring to sort of different things.
For example, a pure sine wave has NO dynamic range, the amplitude of the envelope does not change, the signal’s dynamic range is zero. Conversely, a pure sine in a 5 cycle long Gaussian amplitude envelope like Don Keele uses occupies about 1/3 octave bandwidth but is short enough to test and stress a speaker at very high power without sounding loud at all.
Best,
Tom
I think what was confusing at least for me was the title suggesting this indicated how much power your speakers need (like with music).
What it tells you is how much Voltage it takes to be subjectively pretty loud with a mid bass sine wave.
Depending on the music and recording, it may take 4, 10, 100, 1000 or more times that much (power) to encompass what is in some recordings if played at the same subjective average loudness.
It is like Pano said, digital zero is a point you cannot / should not pass, that level is the loudest / largest Voltage sample that can be recorded on that medium.
For a moment, pretend loudspeakers are dynamically linear, if you put in twice as much power, it puts out the signal 3dB louder and so on (which they do up to between about 1/10 and 1/8 rated power when they begin to compress).
Lets say one had speakers that were 85dB 1w 1m and one sat in an absorptive room (to make it simple) and were 4 meters from each speaker. As this is “stereo” we will take a worst case and say you have a signal on one channel.
Lets say you’re comfortable loud level was only 80dB SPL in that sine wave test.
AS your 4 meters away, due to the inverse square law we know the level is +6dB greater at one meter and we know that at one meter, the speaker was 86dB or 1dB more than 1W or 1.26 Watts.
To get the same SPL reading on a sound level meter using broad band AES pink noise at couple octaves or more wide, it takes an amplifier than can put out 55 Watts (RMS) or twice the Voltage as the signal has a 6dB peak to average ratio.
Similarly with a recording that has a 30dB peak to average ratio, one finds the peaks require 1260 Watts while the average level for the same SPL is still only 1.26 Watts.
Now, +30dB over 80dB sounds like it would be loud, and it would be if it were also a continuous sine wave but if used to reproduce a snare drum or other percussive instrument, it will fall FAR short of the real thing.
It is the very difficulty in capturing and reproducing large dynamic ranges, particularly in a high noise background, that lead to compression and then the low bit MP-3’s so you can hear music when your jogging or with the car top down.
Thankfully one can’t hear instantaneous clipping like traditional “ugh turn it down” clipping, this only saps the dynamics. The only way around this is to have efficient speakers and power or sit very close, or choose your music to fit the system’s capability.
Ultimately, with so many variables in equipment and program material, the ONLY WAY to tell if you have enough Voltage or Power (either view) is to examine what is coming out of your power amplifier with an oscilloscope. If you have one and know how to use it, DO try this, examine the highest peaks and see if any are flat topped.
More often than I would have thought, one finds instantaneous clipping and not just from a power amplifier but any stage can clip, the arrangement of this is “optimizing gain structure”.
Also, it’s worth pointing out that in the recording and engineering areas what we see as dynamic range can be referring to sort of different things.
For example, a pure sine wave has NO dynamic range, the amplitude of the envelope does not change, the signal’s dynamic range is zero. Conversely, a pure sine in a 5 cycle long Gaussian amplitude envelope like Don Keele uses occupies about 1/3 octave bandwidth but is short enough to test and stress a speaker at very high power without sounding loud at all.
Best,
Tom
When I was debating how much power, or rather rail voltage I would need for my HT I hooked up my 2430A and played a low passed and unfiltered signal as loud as I ever would go. Wasn't much so I cranked it up to almost rail voltage, 20V P-P or so, RMS was around 6V I think. A bit too loud and the various squealing noises around the room felt almost louder than the speakers themselves.
Conclusion was that a Class-D amp powered by 24V PSU would do just fine.
Active speakers are a bit harder to calculate for as the voltage of a low or high passed signal can be greater than the input signal. Probably due to phase shift.
(A Hilbert transform of a square wave is probably the best example of this)
Hilbert transform - Wikipedia, the free encyclopedia
That is probably the reason for the previous images showing a greater P-P voltage of pink noise than a sine wave of full amplitude. As all digital to analog converters (sort of) comes with a low pass filter.
Conclusion was that a Class-D amp powered by 24V PSU would do just fine.
Active speakers are a bit harder to calculate for as the voltage of a low or high passed signal can be greater than the input signal. Probably due to phase shift.
(A Hilbert transform of a square wave is probably the best example of this)
Hilbert transform - Wikipedia, the free encyclopedia
That is probably the reason for the previous images showing a greater P-P voltage of pink noise than a sine wave of full amplitude. As all digital to analog converters (sort of) comes with a low pass filter.
Tom, although I get what you are saying I do feel it somewhat misses the point here.
The point is to listen to actual music through your stereo system, this sets the reference level and this is quite an important part. No one is using pink noise or sine waves to determine what is 'loud', this is done with music alone and will obviously vary quite considerably from person to person. This however is another important point though because everyone listens to different types of music and at different levels.
One then plays sine waves through the system at the given volume setting as a way of measuring the voltage required to be able to reproduce digital zero without running into clipping due to voltage limitations. You don't need a scope or fancy average to peak calculations, this is all taken into account before hand when you set the volume control on your hifi with your chosen piece of music.
It is taken for granted that when one person does this and they end up with 2volts, that this would change dramatically if the piece of music was changed to something with a greater peak to average ratio. This however would be unrealistic because that person never listens to that kind of music.
Of course this test doesn't take into consideration the current required by the loudspeakers at the given voltage setting. Some people might be perfectly fine headroom wise when it comes to voltage, but are tripping the protection circuitry due to current limitations in the output stage.
This is taking things to far once again though as the point of the thread was simply to get some sort of rough idea about the amount of power that people are actually using in their hifi systems. There is a lot of talk on the forums surrounding the fact that you need vast amounts of power in order to reproduce peaks faithfully. This is no doubt true if you listen to shuttle launches or extremely dynamic music at high volume levels, but as the poll shows the majority of people don't need this at all.
The point is to listen to actual music through your stereo system, this sets the reference level and this is quite an important part. No one is using pink noise or sine waves to determine what is 'loud', this is done with music alone and will obviously vary quite considerably from person to person. This however is another important point though because everyone listens to different types of music and at different levels.
One then plays sine waves through the system at the given volume setting as a way of measuring the voltage required to be able to reproduce digital zero without running into clipping due to voltage limitations. You don't need a scope or fancy average to peak calculations, this is all taken into account before hand when you set the volume control on your hifi with your chosen piece of music.
It is taken for granted that when one person does this and they end up with 2volts, that this would change dramatically if the piece of music was changed to something with a greater peak to average ratio. This however would be unrealistic because that person never listens to that kind of music.
Of course this test doesn't take into consideration the current required by the loudspeakers at the given voltage setting. Some people might be perfectly fine headroom wise when it comes to voltage, but are tripping the protection circuitry due to current limitations in the output stage.
This is taking things to far once again though as the point of the thread was simply to get some sort of rough idea about the amount of power that people are actually using in their hifi systems. There is a lot of talk on the forums surrounding the fact that you need vast amounts of power in order to reproduce peaks faithfully. This is no doubt true if you listen to shuttle launches or extremely dynamic music at high volume levels, but as the poll shows the majority of people don't need this at all.
For the SW 30Hz to 60Hz sine sweep I got 7.9VAC at 30Hz falling to 7.4VAC at 60Hz. I could actually wach the voltage head south as the frequency went up. cool
good to know my 540w at 4ohm amp is happy. (249.64W peak at 0db)
Pano, per my earlier statement of staying below 60%.
60% of 540w is 324w so at 0db I am under 60%
Not all the inputs into this amp setup are CD players so peak input voltage may vary thus the need for headroom.
sincerely
revb.
good to know my 540w at 4ohm amp is happy. (249.64W peak at 0db)
Pano, per my earlier statement of staying below 60%.
60% of 540w is 324w so at 0db I am under 60%
Not all the inputs into this amp setup are CD players so peak input voltage may vary thus the need for headroom.
sincerely
revb.
Last edited:
Hi 5th I LOVE your movie btw, great sound track too.
My point was that there is a big difference between what “sounds loud” and shows up on a typical sound level meter and the technical requirements needed to reproduce the signal.
With music there is a huge variation in how much dynamic range is present so a generalization is hard to make. I have seen music that was about as compressed as test noise to music that had 20 and even 30 dB difference between the “average (how loud it sounds) and the peak levels.
That means the peaks are 100 to 1000 times more powerful than the average but also there are times where the level is much less too resulting in that average
Earlier in this thread I linked to a free modern “VU” meter and suggested people examine their music files and see what they have in them.
It shows “level” from several different perspectives, the slowest is more like our subjective impression of loudness and the fastest being what is required to faithfully reproduce the actual signal.
Here it is again;
ORBAN Loudness Meter
When you examine a musical track, what you can see is there is an average or VU and several other views that show increasing amounts of the actual dynamics.
Understand the VU meter came about when things clipped more gradually with full awareness that you can’t hear distortion or even clipping if short. It’s time constant more or less reflects ones impression of loudness.
When showing the shortest peaks, these are what the actual signal contains, these very short events don’t sound as loud as they are because of our hearing, but these dynamics are part of what make those sounds seem real.
Lastly much better than thinking you have no problem with instantaneous clipping because you can’t hear anything, the only way to actually KNOW if you have enough power is to listen to music at a level of your choosing AND examining the Voltage peaks going to the speaker with an oscilloscope.
When in doubt, remember she said eckto gamut
Dynamically,
Ruby R
My point was that there is a big difference between what “sounds loud” and shows up on a typical sound level meter and the technical requirements needed to reproduce the signal.
With music there is a huge variation in how much dynamic range is present so a generalization is hard to make. I have seen music that was about as compressed as test noise to music that had 20 and even 30 dB difference between the “average (how loud it sounds) and the peak levels.
That means the peaks are 100 to 1000 times more powerful than the average but also there are times where the level is much less too resulting in that average
Earlier in this thread I linked to a free modern “VU” meter and suggested people examine their music files and see what they have in them.
It shows “level” from several different perspectives, the slowest is more like our subjective impression of loudness and the fastest being what is required to faithfully reproduce the actual signal.
Here it is again;
ORBAN Loudness Meter
When you examine a musical track, what you can see is there is an average or VU and several other views that show increasing amounts of the actual dynamics.
Understand the VU meter came about when things clipped more gradually with full awareness that you can’t hear distortion or even clipping if short. It’s time constant more or less reflects ones impression of loudness.
When showing the shortest peaks, these are what the actual signal contains, these very short events don’t sound as loud as they are because of our hearing, but these dynamics are part of what make those sounds seem real.
Lastly much better than thinking you have no problem with instantaneous clipping because you can’t hear anything, the only way to actually KNOW if you have enough power is to listen to music at a level of your choosing AND examining the Voltage peaks going to the speaker with an oscilloscope.
When in doubt, remember she said eckto gamut
Dynamically,
Ruby R
OK I've done the test. I'd never normally listen this loud, but it is conceivable for a party (I used to play much louder at parties in my youth).
Measured voltage on the 120hz test tone was 4.0V
test track was Telegraph Road from Love over gold which has about 12db range from average to peak if I understand the output of ocenaudio.
Speakers are probably around the 87-88db 2.38V 1M mark. They are nominally 4 ohms.
Listening room is approx 9M X 5M and listening distance is about 3M.
Tony.
Measured voltage on the 120hz test tone was 4.0V
test track was Telegraph Road from Love over gold which has about 12db range from average to peak if I understand the output of ocenaudio.
Speakers are probably around the 87-88db 2.38V 1M mark. They are nominally 4 ohms.
Listening room is approx 9M X 5M and listening distance is about 3M.
Tony.
Attachments
As I said before, I do agree with what you are saying and understand it, but the methods used here will suffice completely for the intended end goal - to see if people actually have enough power, or voltage headroom, for the majority of their listening requirements.
This is indeed true, but it is no different (on the whole) to what Pano has suggested.
I am wanting to build a quad of class A power amplifiers to drive the mids and tweeters in my setup and I wanted to see what voltage rails would be adequate for very loud listening levels. Naturally the lower the supply rails, the lower the device dissipation for a given amount of standing current so there's good incentive to keep them low. Due to the nature of the system, scoping was absolutely necessary. The system is completely active and uses a wave-guide on the tweeter. This boosts the sensitivity of the tweeters lower range, but does little at the frequency extremes. Obviously music doesn't contain much beyond 10khz and if I'd used sine waves as the stimulus I'd need much higher voltage rails just so I can reproduce the full scale output @ 20khz. Scoping the output with many different types of music was necessary to see what would actually be required, but this is working in reverse.
I could have chosen a listening level as per Pano's suggestions, then played sine waves through the system. As the signal level being sent to the tweeter actually increases as frequency increases, the amplifier driving the tweeter would reach its amplitude maximum at around 20khz. If I chose this to set the voltage rail for the class A amplifiers I would know without a shadow of a doubt that I'd never clip the tweeter amplifier, but it would be completely overkill. Using the scope actually arrived at a lower figure then using the sine waves because music never contains full scale output at 20khz. This you could say is a little bit of a gamble, what if a piece of music comes along that does have a lot of top octave energy? Well then my tweeter amplifier might clip! If I'd built it as per the sine wave measurement then it probably would not.
Both methods have their uses and I guess it really boils down to whether or not you understand what it is you are doing and what the information you're getting tells you. I guess from your position you are probably trying to offer caution towards what people are doing in an attempt to say that just because the test you did today shows you wont clip the amplifier doesn't mean that in the future another piece of music will be clip free too. Of course if they leave the volume control in exactly the same place it will be clip free.
This is indeed true, but it is no different (on the whole) to what Pano has suggested.
I am wanting to build a quad of class A power amplifiers to drive the mids and tweeters in my setup and I wanted to see what voltage rails would be adequate for very loud listening levels. Naturally the lower the supply rails, the lower the device dissipation for a given amount of standing current so there's good incentive to keep them low. Due to the nature of the system, scoping was absolutely necessary. The system is completely active and uses a wave-guide on the tweeter. This boosts the sensitivity of the tweeters lower range, but does little at the frequency extremes. Obviously music doesn't contain much beyond 10khz and if I'd used sine waves as the stimulus I'd need much higher voltage rails just so I can reproduce the full scale output @ 20khz. Scoping the output with many different types of music was necessary to see what would actually be required, but this is working in reverse.
I could have chosen a listening level as per Pano's suggestions, then played sine waves through the system. As the signal level being sent to the tweeter actually increases as frequency increases, the amplifier driving the tweeter would reach its amplitude maximum at around 20khz. If I chose this to set the voltage rail for the class A amplifiers I would know without a shadow of a doubt that I'd never clip the tweeter amplifier, but it would be completely overkill. Using the scope actually arrived at a lower figure then using the sine waves because music never contains full scale output at 20khz. This you could say is a little bit of a gamble, what if a piece of music comes along that does have a lot of top octave energy? Well then my tweeter amplifier might clip! If I'd built it as per the sine wave measurement then it probably would not.
Both methods have their uses and I guess it really boils down to whether or not you understand what it is you are doing and what the information you're getting tells you. I guess from your position you are probably trying to offer caution towards what people are doing in an attempt to say that just because the test you did today shows you wont clip the amplifier doesn't mean that in the future another piece of music will be clip free too. Of course if they leave the volume control in exactly the same place it will be clip free.
BiAmp (Bi-Amplification - Not Quite Magic, But Close) - Part 1
please read after 1.3.
this is a good explanation for why my findings were so low
and far from reality.
a 4x with passive systems seems correct.
at least from what i hear at my system.
please read after 1.3.
this is a good explanation for why my findings were so low
and far from reality.
a 4x with passive systems seems correct.
at least from what i hear at my system.
As I mentioned earlier, 16 bit = 65536 discrete steps. The Philips disc also say this
"Level Definition"
The reference level of 0db refers to a sinewave whose most positive and negative value corresponds to +32767 quantisation steps and whose negative value corresponds to -32767 quantisation steps
So... and this is something I have never considered... is that saying the sinewave is "biased
I honestly don't know
You forgot to count 0 V as a level. So you have 32767 positive voltage levels, 32767 negative voltage levels, and zero. So you have 65535 distinct voltage levels and you throw one away.
On a theoretical basis nothing says you can't use all 65536 possible values, but then you would not have a specific quantization for 0.0 V (while maintaining |V+(max)|==|V-(max)| ). And while an absolute zero is not required, it's just inelegant not having one.
-bill
Last edited:
I think the reason for throwing away -32768 is just related to two's complement arithmetic. Calculations with negative numbers become very simple when you use two's complement number representations, but two's complement coding always has one value more on the negative than on the positive side.