Quoted from post
http://www.diyaudio.com/forums/solid-state/240712-cfa-topology-audio-amplifiers-43.html#post3619627
"To handle an uncompressed (or lightly compressed) recording at decent sound pressure levels then huge amounts of power are required and this is expensive. So people accept clipping as a consequence of this. Having good clipping behaviour is a way of mitigating the issue."
At the light of many data I collected on sound levels from many sources, I disagree.
Note : below, pk/avg stands for ratio peak/average.
Famous in France among diyers for his listening and HC dedicated room, Roland Delacroix stated that the highest dynamics he ever exceptionnaly encountered in a record was 68 dB(pk/avg = 34 dB) and told it is almost impossible to handle. Even 52-55 dB (pk/ag = 27 dB) are extremely rare.
Source :
dolby lake process : Enceintes Hi-Fi - Cinetson - Hifi et Homecinema
Pano who is a moderator of this forum has found that the best CDs from the point of view of dynamics have a pk/avg about 22 dB or more.
Source :
http://www.diyaudio.com/forums/soli...atts-im-using-during-music-2.html#post3166044
So we can state that a system able to reproduce dynamics of almost all existing records needs to reproduce a pk/avg of 25 dB without distorsion.
High but still decent average sound levels may rely between 75 dB (at home) and 85 dB (at the recording studio) SPL. This means that the loudspeakers should deliver peak levels of 100 dB to 110 dB at the listening position.
An often quoted target is 105 dB SPL peak.
To my experience, this is far more than most people need for everyday listening.
Assuming at the listening position for a two channel stereo system, a loss of 3 dB from the nominal sensitivity of the loudspeakers, we ended, if they have a nominal sensitivity of 90 dB SPL / 2.83 V at 1m (1 w in 8 Ohm) with the following levels :
81 dB 0.25 W
87 dB 1 W
93 dB 4 W
99 dB 16 W
105 dB 64 W(peak)
This is for a typical two channel stereo system and for SPLs somewhat higher than usual. It does not represent huge amounts of power and there is no reason why an adequately dimensioned amp for this system should ever clip.
http://www.diyaudio.com/forums/solid-state/240712-cfa-topology-audio-amplifiers-43.html#post3619627
"To handle an uncompressed (or lightly compressed) recording at decent sound pressure levels then huge amounts of power are required and this is expensive. So people accept clipping as a consequence of this. Having good clipping behaviour is a way of mitigating the issue."
At the light of many data I collected on sound levels from many sources, I disagree.
Note : below, pk/avg stands for ratio peak/average.
Famous in France among diyers for his listening and HC dedicated room, Roland Delacroix stated that the highest dynamics he ever exceptionnaly encountered in a record was 68 dB(pk/avg = 34 dB) and told it is almost impossible to handle. Even 52-55 dB (pk/ag = 27 dB) are extremely rare.
Source :
dolby lake process : Enceintes Hi-Fi - Cinetson - Hifi et Homecinema
Pano who is a moderator of this forum has found that the best CDs from the point of view of dynamics have a pk/avg about 22 dB or more.
Source :
http://www.diyaudio.com/forums/soli...atts-im-using-during-music-2.html#post3166044
So we can state that a system able to reproduce dynamics of almost all existing records needs to reproduce a pk/avg of 25 dB without distorsion.
High but still decent average sound levels may rely between 75 dB (at home) and 85 dB (at the recording studio) SPL. This means that the loudspeakers should deliver peak levels of 100 dB to 110 dB at the listening position.
An often quoted target is 105 dB SPL peak.
To my experience, this is far more than most people need for everyday listening.
Assuming at the listening position for a two channel stereo system, a loss of 3 dB from the nominal sensitivity of the loudspeakers, we ended, if they have a nominal sensitivity of 90 dB SPL / 2.83 V at 1m (1 w in 8 Ohm) with the following levels :
81 dB 0.25 W
87 dB 1 W
93 dB 4 W
99 dB 16 W
105 dB 64 W(peak)
This is for a typical two channel stereo system and for SPLs somewhat higher than usual. It does not represent huge amounts of power and there is no reason why an adequately dimensioned amp for this system should ever clip.
Member
Joined 2009
Paid Member
I figure you need to ask yourself how often is too often for clipping. A music signal almost looks random, a lot of peaks at low level, very few at high level, vanishingly few at very high level. If it clips rarely - do you care ? I don't. Many compromises are involved in music playback and I am quite happy not to be running 'megawatt' amplifiers.
This is an important question. There are 2 questions/answers and it depends on the type of music.
1 Is it audible? If you clip a kick drum, most people can tell the difference but most don't find it objectionable (apologies to yus Golden Pinnae who object if a gnat sits on your mains cable. There is a LOT of modern music (?) where if the amp clips 50% of the time, even the best ears in the business can't tell the difference.
http://www.aes.org/e-lib/browse.cfm?elib=3793
2 Is it nasty? Unlike the kick drum, if you clip a good recording of a GOOD unaccompanied choir of about 12-20, this is certainly objectionable. Also holds true for good piano.
The choir can clip the best 16b recorders & microphones if recorded simply.
Most commercial recordings don't come anywhere near this cos evil compression is applied. That's for classical. For pop ...
forr, we surely have to consider speaker efficiency in this discussion. 90 dB seems a typical figure, although I note in a number of tests, many speakers fall short by 1 or 2dB (might be measurement methodology - I dont know).
I did read an article where it was claimed (by Anthony Michelson of Musical Fidelity) that very high powers were required to produce realistic, undistorted levels for orchestral and piano music. I guess here he would be talking about speakers in the 86-90 dB sensitivry range. The output power quoted was 500W to 1 000W - peak of course. This puts us in B&W 800 or Avalon territory. MF also produced a chart to show amplifier power needed to produce a given SPL vs speaker efficiency.
But, back to my original point, we should assess output power on 3 requirements - desired max listening levels in the home, peak music levels on the material and speaker efficiency. The other factor is the listening space.
BTW I am currently using a 15W class A amp, a 100W AB and a 180W (really 250W) class AB driving B&W 703's and on occasion a small pair of very fine Taiwan book shelf speakers. But, I don't listen at realistic orchestral levels (well maybe when my partner is out and I know the neighbors are not around)
I did read an article where it was claimed (by Anthony Michelson of Musical Fidelity) that very high powers were required to produce realistic, undistorted levels for orchestral and piano music. I guess here he would be talking about speakers in the 86-90 dB sensitivry range. The output power quoted was 500W to 1 000W - peak of course. This puts us in B&W 800 or Avalon territory. MF also produced a chart to show amplifier power needed to produce a given SPL vs speaker efficiency.
But, back to my original point, we should assess output power on 3 requirements - desired max listening levels in the home, peak music levels on the material and speaker efficiency. The other factor is the listening space.
BTW I am currently using a 15W class A amp, a 100W AB and a 180W (really 250W) class AB driving B&W 703's and on occasion a small pair of very fine Taiwan book shelf speakers. But, I don't listen at realistic orchestral levels (well maybe when my partner is out and I know the neighbors are not around
)
Last edited:
Hi Bonsai,
I notice you built a 15 watter Class A and a no. of Class AB amps, if you compare them can you distinctively distinguish their differences(*) other than power output, I mean sound reproduction. * I mean Class A sound seems smoother(more liquid) and "more" effortless than Class AB. Any other differences you want to write about?
I notice you built a 15 watter Class A and a no. of Class AB amps, if you compare them can you distinctively distinguish their differences(*) other than power output, I mean sound reproduction. * I mean Class A sound seems smoother(more liquid) and "more" effortless than Class AB. Any other differences you want to write about?
Some interesting links
Amplifier Power Required Calculator
for 2m from the speakers, 100 dB SPL and allowing for 3 dB of headroom with 89 dB speakers, I need 100 Watts
Here is a discussion at Audioholics
http://www.audioholics.com/loudspeaker-design/loudspeaker-sensitivity
Amplifier Power Required Calculator
for 2m from the speakers, 100 dB SPL and allowing for 3 dB of headroom with 89 dB speakers, I need 100 Watts
Here is a discussion at Audioholics
http://www.audioholics.com/loudspeaker-design/loudspeaker-sensitivity
CDs with high dynamic range
Can you post a TXT file with your examples?
This was of great importance to me in my previous life and I had my own lists of commercial CDs. Sadly, I no longer have my list
The only ones I sorta remember are Simon Rattle's recording of Britten's War Requiem and a very old Holst Planets.
Very little pop appears on this list. Most of of my high dynamic range examples are my own unprocessed recordings.
forr, none of those links seem to have lists of the CDs with high dynamic range. I believe you have your own list.
Can you post a TXT file with your examples?
This was of great importance to me in my previous life and I had my own lists of commercial CDs. Sadly, I no longer have my list
The only ones I sorta remember are Simon Rattle's recording of Britten's War Requiem and a very old Holst Planets.
Very little pop appears on this list. Most of of my high dynamic range examples are my own unprocessed recordings.
Records are more often clipped than amplifiers do.
There are many freeware tools available for examining the content of digital records from different points of view. They are very informative. It is quite amazing how many records reach FSD. Nobody guesses they contain clipped moments, they do not sound unpleasant at all.
I'll give some references of softwares and records later.
There are many freeware tools available for examining the content of digital records from different points of view. They are very informative. It is quite amazing how many records reach FSD. Nobody guesses they contain clipped moments, they do not sound unpleasant at all.
I'll give some references of softwares and records later.
As I said, and investigated formally in the papers I quoted, there is music (??) where even the best ears in the business find it hard to tell if the amp clips 50% of the time.
BTW, that's the music (??) clipped 50% of the time .. NOT that true golden pinnae can only tell 50% of the time.
An example of the choir effect requiring huge peak capability even at modest average levels I mentioned on a commercial recording is the Preston/Hogwood Messiah with the AAM and Christchurch Cathedral Choir.
If you record such choirs yourself, you'll find it very common. Unlike most pop, overload here is obvious and unpleasant.
I look forward to your list of CDs with high dynamic range.
All of Kavi's recordings (Water Lilley Records) are uncompressed recordings.... LP and CD.
On the power requirments ---> the size/volume of the listening room (and if it is live or dead or in the middle) is needed to be factored into the power needed. There is a standard European size and a slightly larger american room size. Mine is larger than most common room sizes.
I need about 250W/8 minimum for 105-110dB spl. BUT, when driving bass speakers that go down to 20Hz, a lot more power is needed down low (4 each 15 inch bass drivers and 2 each 18 inch subs). It isnt loud at freq below 100hz or less without higher power drivers/amps. Gots ta move a lot of air to make a difference.
Although each individual can decide how much clipping they can tolerate - I really just need to know the amp power for Not clipping.... then work backwards from there re-clipping percentage.
Thx-RNMarsh
Last edited:
ttan,
yes I have built 4 different amps covering VFA, CFA class A and class AB.
I believe you really have to match your amp to the music you listen too in most cases (maybe in cost no object systems you can relax this a bit).
I use my sx-Amp, which is class A 15 W, for classical and acoustic, and the nx and e-Amps for Jazz and heavier music.
Be aware, that these intepretations of 'sound' are highly subjective and 'sound is in the ear of the listner'. What sounds good to me, may not neccessarily be your cup of tea.
The sx-Amp certainly has a very relaxed sound to my ears, and is very smooth, but let me add the differences in sound between the various amps is not night and day - it should not be if all the right technical boxes are ticked of course.
Here is a link to the nx-Amp write up - in the back, I discuss my perceptions of the sound The Ovation nx-Amplifier V2.0 and here is the sx-Amp write up with comments on the sound as well in the last section The Ovation sx-Amplifer V2.0
Last edited:
If the "clipping" is built into the recording, then THE AMP IS NOT CLIPPING, when it is played (unless it's turned up too far and is actually clipping the clipping, of course).
So there would be no reason to expect anyone to notice, since the amp does not clip, and thus no nasty artifacts are produced.
"Clipping" of the amplifier would, by definition, mean that the output signal power level had attempted to exceed the true maximum available output power of the amplifier, because the reservoir capacitance and/or the power rail voltage were/was too small for the attempted output level. The signal voltage comes up, the ripple voltage dips down, and the voltage between the amplifier's power rail input and its audio output is squeezed in between the two. If it gets squeezed down to the "Vclip" minimum voltage level, then ripple-shaped chunks start getting gouged out of the output signal waveform. (The Vclip voltage is exactly like the dropout voltage of a regulator.) Squeeze even more and the entire top (and/or bottom) of the output waveform could be sheared off.
The signal that is heard from the speakers is CURRENT that comes rather-directly from the reservoir capacitors in the power supply. (SEE the attached image! It's what the cap currents and the speaker voltage are doing during a snippet from the intro to the song "Highway to Hell", by AC/DC.)
Drawing music current from the caps causes their voltage to fall, between charging pulses. That is what causes the ripple voltage.
The idea is to have enough reservoir capacitance that the maximum allowed music current won't ever draw the caps' voltage down more than the desired ripple amplitude, before the caps can recharge.
A properly-designed amplifier can never clip, and can never run out of current for even the lowest bass tones, between charging pulses, IFF it doesn't attempt to exceed its RATED (i.e. designed) maximum output power level.
Many amplifier power supplies are designed "wrong" (according to me). Their reservoir capacitance value is often designed to accommodate only the maximum RMS output current of a single sine wave. But in order to be able to handle ANY signal waveform, and any low frequency, the total capacitance value should be designed to be able to accommodate constant DC current that is at the PEAK (not RMS) output level that corresponds to the rated maximum output power.
When an amplifier does clip, the sharp edges that are created will be composed of a spray of very high frequencies (narrow in the time domain = wide in the frequency domain). The edges are sharp, and at high power, and they stay sharp because they are created just before they leave the amplifier and there's not much there to filter them. (Tweeters beware.)
One difference between real clipping, of the amplifier, and "clipping" that is already in the source material, is that the sharp edges are no longer sharp, in the source material version. The high-frequency content that is above the audio range, and most of the destructive energy, has been removed, by low-pass filtering that is built in throughout the audio recording production chain.
Real clipping can blow the ribbon tweeter fuses in my Magnepan MG-3.6/R speakers. The "clipping" that is already in the recording cannot.
Another difference is that during real clipping, the ripple voltage amplitude is very large, and past its design limit, and the reservoir caps are being exhausted and the rail voltage is dipping very low, whereas the source-material "clipping" would normally be playing when the amplifier is doing just fine.
Also, during real clipping, there will be very high frequencies included in the power rail ripple, since the output signal current is causing the ripple, and the clipping is making sharp edges in the output waveform, i.e. high frequencies.
Amplifiers usually have very good PSRR (power supply rejection ratio), especially at lower frequencies, which enables them to not distort very much, even though their power rails have ripple on them. But always some distortion occurs, and very large ripple amplitudes will cause more distortion than small-amplitude ripple. And very high frequencies in the ripple would cause even more distortion, because the PSRR usually gets much worse, fairly quickly, as frequency increases.
There is also a hidden positive feedback loop for high frequencies, through the power rail, in most transistor-based amplifiers. So the HF on the rail could also cause ringing or bursts of HF oscillation, especially if the HF bypass caps haven't been implemented properly (i.e. extremely close to each output device).
I guess that's enough to make me believe that real clipping could be audible, at least. Sorry for blathering-on for so long, about all of that. (And I hope that Richard will enlighten me if I have muddled the concepts too much, while thinking out loud, here.)
Cheers,
Tom
The electrolytic capacitors ARE THE SIGNAL PATH:
Attachments
Last edited:
In the JAES reference we point out the existence of 'music' (??) where it makes no difference (ie undetectable even to true golden pinnae) if the amplifier is clipped or not.
Far be it for a mere speaker and mike designer to imply that such recordings are already clipped or have so much distortion that amplifier clipping makes no audible difference.
_____________
I used to advise anyone designing speakers or mikes to go out & record some live music themselves. Whether they used cheap Sony ECMs and a Walkman or Soundfield Mike & PCM-F1, this would change their whole perception and philosophy of recording & reproduction.
Sadly, this advice is no longer valid. This Millenium when MUCH better recorders & microphones are inexpensively available to the unwashed masses, they all just want to emulate their idols, with Jurassic big name mikes and the latest clippers/compressors and other evil stuff.
What's happened to Dynamic Range that us old fogeys were so intent on pursuing? We thought this was a vital part of music making ... even though it made life hard for mike, speaker & amp makers.
You know DYNAMIC RANGE; soft bits contrasting with loud bits
Is it ALL just mezzo forte con belto these days?
As Crowhurst pointed out in '57 (60 years ago) is that the NFB error signal goes through the roof as the output fails to keep up with the input.
This then allows one to clip the stages of the amplifier, usually with the effect of upsetting some coupling or DC bias or DC supply capacitors.
I once had a circuit using a quad op amp in which one amp clipped hard enough to upset all the other amps to the point of circuit failure - and this in an industrial application, not audio.
Do not trust this kind of theorical calculators.
This one seems to be based on a square law attenuation of SPL with distance. It does not reflect what happens in domestic conditions where there are two loudspeakers and a great amount of reverberation.
An rule of thumb sometimes given for a two channels stereo system is that the SPL at 3 m for 2.83 V is about -3 dB less than the nominal sensitivity at 1 m for the same voltage of the loudspeakers.
Using pink noise and a sonometer (C weighting) is enlightning.
The discussion here aim to size the power of amplifiers and to avoid over- or under- estimate the real needs.
What is important is to avoid amp clipping. So, since long, I measured peak voltages and currents (Metrix and Fluke meters) delivered by my amps with the most demanding records I could find.
My amps are rougly rated above 10 dB of the maximal voltages and currents they have ever to deliver.
Here is refered a handy little tool to investigate the power distribution in records :
http://www.diyaudio.com/forums/soli...equency-distributions-wanted.html#post2871453
Other programs of the same category (there may be many others) which allow the analysis of spectral and power distributions as well as dynamics of what you listen to :
Ocenaudio
TT-DR offline 1.4
Wave Spectra
Last edited:
In the JAES reference we point out the existence of 'music' (??) where it makes no difference (ie undetectable even to true golden pinnae) if the amplifier is clipped or not.
Far be it for a mere speaker and mike designer to imply that such recordings are already clipped or have so much distortion that amplifier clipping makes no audible difference.
_____________
I used to advise anyone designing speakers or mikes to go out & record some live music themselves. Whether they used cheap Sony ECMs and a Walkman or Soundfield Mike & PCM-F1, this would change their whole perception and philosophy of recording & reproduction.
Sadly, this advice is no longer valid. This Millenium when MUCH better recorders & microphones are inexpensively available to the unwashed masses, they all just want to emulate their idols, with Jurassic big name mikes and the latest clippers/compressors and other evil stuff.
What's happened to Dynamic Range that us old fogeys were so intent on pursuing? We thought this was a vital part of music making ... even though it made life hard for mike, speaker & amp makers.
You know DYNAMIC RANGE; soft bits contrasting with loud bits
It appears that I misunderstood the point and have spoken out of context.
The advice you mentioned is still valid, even if it is unheeded or unappreciated, or even deemed unneeded by many in the newer generations.
Tragically, as in so many other fields, and culture in general, the vast majority will never realize what they are missing. I fear that too much will be lost. It's profoundly saddening to witness the decline, and to realize that there is likely no one who will pick up your torch and continue your upward climb.
It's plain math - NFB works by using gain to keep the error signal close to "zero". One the output hits the rail, but the input keeps rising (or falling) the error signal gets "large" very quickly.
Now if you've got more gain than headroom in the internal stages of the amplifier, they are all going to clip shortly after the output stage clips.
What happens next is unlikely to be pleasant.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.