• Welcome to the all-new HUG. All content has been converted from the old system, and over the next days we will re-style HUG in a more familiar way.

"Fluidity"? >>> clipping & amp sonics

Status
Not open for further replies.

A.S.

Administrator
Staff member
Clarification?

Clarification?

It would help to be consistent if you ripped the track and placed in a waveform editor (like Audition) so we can get a sense of perspective here. Most of our readers are non-technical, and whilst technicians can probably interpret two different displays, we cannot assume that.

Which track is it exactly?
 

mhennessy

Member
'From this Moment On'

'From this Moment On'

It would help to be consistent if you ripped the track and placed in a waveform editor (like Audition) so we can get a sense of perspective here. Most of our readers are non-technical, and whilst technicians can probably interpret two different displays, we cannot assume that.
Yes, I've done that, but as I say, I don't seem to be able to upload pictures. I created a Flickr account and put the one photo on as a test, but then I needed to get some sleep!

Is this the best way to add images to this forum - are uploads restricted to moderators only?

Which track is it exactly?
"From This Moment On", from the album of the same name
All the best,

Mark
 

A.S.

Administrator
Staff member
Yes, photos are restricted. We just don't have the manpower to increase the moderation load.

If you wish to use an external photo/image store, that's fine. I may have this album on CD. I'd really like to see the waveform of the entire track.
 

Pluto

New member
Exact disc in question?

Exact disc in question?

... take a look at a quick grab of the Diana Krall track I keep mentioning ("From This Moment On", from the album of the same name). Notice the large positive-going "spikes"; that's the brass instruments that are very forward in the mix - I'd guess that the peaks are roughly 12dB above the rest. In power terms, 12dB is 16 times greater!
I've just borrowed this disc from a neighbour and when you look at the entire 3 minutes, this is what you get

(hover your mouse over the small image to see a magnified view)

You must be registered for see images attach


Superficially, this looks like a heavily clipped song, a subject we have been talking about quite a lot lately. But… let's zoom in to roughly that area that Mark has shown on his 'scope photo. In this image, you are looking at about 30mS of audio…

You must be registered for see images attach


and what you are seeing are individual 'spikes' in the waveform that make up the brass punches. You can see that these spikes are not flat-topped, the true give-away of clipping, but have pointed tops which indicates that the actual gain of the system has been adjusted, automatically, by a device known as a limiter, so that the very peak of those spikes are at the maximum possible level. This is very clever mastering because all the detail in between the spikes is preserved, intact.

So, in order to reproduce this disc well, as Mark has already stated, it is necessary to have an amplifier with sufficient power to reproduce the absolute maximum level* of which digital audio is capable, cleanly and without clipping. Naturally, this power requirement is dependent on how loud you listen, but is easily worked out.

I hope that Mark will soon put his scope across his loudspeaker terminals and tell us what actual voltages are necessary to play this track at a decent level.

* unlike analogue audio, digital has an absolute maximum level that exists when the sample in question consists entirely of ones i.e. in the case of 16 bit audio, a value of 65535. Sixteen bits can represent a number no higher than this.
 
Attachments only viewable to members

A.S.

Administrator
Staff member
Caution over observations

Caution over observations

Now this is very odd indeed. In my post #81, and reply #82, I mentioned that we were unable to see from a tiny oscilloscope snapshot the totality of the entire track waveform. Pluto has ripped the track and provided that now, using the same software I used to make the example in my post #73.

Taking Pluto's waveform of the CD rip at face value, what we witness is the Krall track, first mentioned in post # 63, to be significantly clipped. It has the same generic tell-tale waveform scything as my example in post 73. This must have occurred during the recording and/or mastering process. It is nothing to do with the CD player or the ripping process.

Now, we are in serious danger of getting ourselves tied in knots. Is the track in question clipped or not? If it is clipped, how does this sit with the observations in post #63 where the listener was aware of a change in the nature of sound when the (clipped?) track was played through a low power amplifier, but not aware of the clipped recording when playing it through a high power amplifier? Surely, if the source track is demonstrable (visually) clipped, playing it through a powerful amplifier can't somehow 'un-clip' it can it? Or can it to the human ear? Maybe it can.

Do we have a two step process in action: a clipped music track replayed by a clipped amplifier sounds distinctively different to a clipped music track being played by a big amplifier, not itself clipping?
 

mhennessy

Member
Hi Pluto,

Just quickly before I head off home :)

Ignore the vertical scale. I got that signal directly from the CD player - in fact from the headphone socket on the front panel - it was just easier to plug up quickly!

That picture was taken in the workshop late last night. I'll hopefully get the chance to set up an oscilloscope in the music room later this evening.

Cheers,

Mark
 

mhennessy

Member
Hot mastering

Hot mastering

Taking Pluto's waveform rip at face value, what we see is the Krall track, first mentioned in post # YY, to be significantly clipped. It has the same generic tell-tale waveform scything as my example in post YY. This must have occurred during the recording and/or mastering process. It is nothing to do with the CD player or the ripping process.
When I ripped it with Exact Audio Copy, it reported the highest peak to be within a fraction of a dB of full-scale, but no samples actually reached full-scale.

When you observe with an analogue 'scope - often the best way to spot clipping because of the bright-spot effect - there's no obvious sign. I think that it's been mixed and mastered to be fairly "hot" in terms of getting close to 0dBFS - like pretty much any contemporary CD released in the last 20 years - but I don't think it's been hard-clipped during that process. I wouldn't mind betting that compression and soft limiting has been artistically applied to individual stems during the recording and mix-down, and perhaps some careful peak-level normalisation across the whole track at some point - but the result is still dynamic and punchy. Have you had a chance to listen to it yet?

All that said, supposing those brass peaks are clipped, we have no way of knowing how clipped they are. They might have been be 5 times the size for all we know. Or, they might be just 1% bigger in reality. But importantly, for our purposes, it actually doesn't matter. We have the CD recording which is a reference (A) - if we go on to clip it further (B), then we have two different things to AB compare.

Last night before going to sleep, I applied a hard limit at -3dBFS, and even on the workshop setup, I was able to hear a difference when doing an AB test (easy to do - I just applied the clipping to random parts of a ~50 second excerpt, then put something else on the screen so I couldn't see whether I was hearing original or unclipped). The effect was more subtle than the piano excerpt, but then so was the limit applied, I think (what level did that use?). If I get time, I'll repeat the experiment upstairs in the music room using a CD created from the .WAV file. Pluto - fancy trying to so something similar?

Just to recap what I've said previously:

All I'm trying to demonstrate is that mild clipping can alter the sound in subtle ways. It doesn't necessarily make it worse. Clipping is very common in the programme material anyway, often for artistic or commercial reasons. But, it can be one of the mechanisms that contributes to audible differences between different amplifiers, and it's more pervasive than we might imagine - perhaps not in everyday domestic listening, but I've seen (or heard!) it a lot in hi-fi showrooms and at work.

(On more than one occasion I've seen audio operators criticise loudspeakers because they've been over-driven at high levels by grossly clipping amplifiers - as an engineer I immediately recognise the sound as clipping, but have to put a 'scope across the terminals to prove it to them.)

Cheers,

Mark
 

anonymous

New member
Which is which?

Which is which?

But as to which one is clipped and which isn't, is a matter of objective fact.
Not to overstep the questions about the Diane Krall track, but which is the clipped section from your original example?

It's funny. You characterized the clipped version as "grossly clipped" but it really doesn't strike me that there is a "night and day" difference between them. And, for someone like myself, who usually listens to music at lower volumes to preserve both my hearing and the harmony of the household, it seems likely it'd be even less apparent.
 

A.S.

Administrator
Staff member
Not to overstep the questions about the Diane Krall track, but which is the clipped section from your original example?

It's funny. You characterized the clipped version as "grossly clipped" but it really doesn't strike me that there is a "night and day" difference between them. And, for someone like myself, who usually listens to music at lower volumes to preserve both my hearing and the harmony of the household, it seems likely it'd be even less apparent.
Well, Anonymous, I thank you for stepping forward. Well done you.

'Grossly clipped' is evident from a visual examination of the waveform. That would be a really serious extreme, although no doubt, an even more aggressive clipping could be applied, and even then may only be barely detectable, with this particular music.

I'll email you privately, as you certainly deserve to know.

*{Moderator's comment: Sorry. Not permitted any more here: see >> here <<}
 

A.S.

Administrator
Staff member
When I ripped it with Exact Audio Copy, it reported the highest peak to be within a fraction of a dB of full-scale, but no samples actually reached full-scale.

When you observe with an analogue 'scope - often the best way to spot clipping because of the bright-spot effect - there's no obvious sign. I think that it's been mixed and mastered to be fairly "hot" in terms of getting close to 0dBFS - like pretty much any contemporary CD released in the last 20 years - but I don't think it's been hard-clipped during that process. I wouldn't mind betting that compression and soft limiting has been artistically applied to individual stems during the recording and mix-down, and perhaps some careful peak-level normalisation across the whole track at some point - but the result is still dynamic and punchy. Have you had a chance to listen to it yet?

All that said, supposing those brass peaks are clipped, we have no way of knowing how clipped they are. They might have been be 5 times the size for all we know. Or, they might be just 1% bigger in reality. But importantly, for our purposes, it actually doesn't matter. We have the CD recording which is a reference (A) - if we go on to clip it further (B), then we have two different things to AB compare.

Last night before going to sleep, I applied a hard limit at -3dBFS, and even on the workshop setup, I was able to hear a difference when doing an AB test (easy to do - I just applied the clipping to random parts of a ~50 second excerpt, then put something else on the screen so I couldn't see whether I was hearing original or unclipped). The effect was more subtle than the piano excerpt, but then so was the limit applied, I think (what level did that use?). If I get time, I'll repeat the experiment upstairs in the music room using a CD created from the .WAV file. Pluto - fancy trying to so something similar?

Just to recap what I've said previously:

All I'm trying to demonstrate is that mild clipping can alter the sound in subtle ways. It doesn't necessarily make it worse. Clipping is very common in the programme material anyway, often for artistic or commercial reasons. But, it can be one of the mechanisms that contributes to audible differences between different amplifiers, and it's more pervasive than we might imagine - perhaps not in everyday domestic listening, but I've seen (or heard!) it a lot in hi-fi showrooms and at work....
My mistake. Because of the Y scale, visually, it looked as if there was black (no signal) above and below the waveform that was not being used i.e. the peaks were not as 0dB. In fact, properly examining the scale shows that the signal is, obviously now, very close to the end stops. Pluto has used a different Y scale to me. Another potential confusion for the reader.

We could pursue this for weeks - it's a fascinating subject - but we can't. The contributing audience is too small and we cannot talk technical amongst ourselves. Could we draw this to a conclusion for the benefit of all, supported by some real world audio examples please?

What we need to know is, can we hear sound clip evidence (not technical talk about dBs and whatever) that supports your very bold statement:

... it's important to realise that most amplifiers are clipping most of the time. OK, not for background music, but when one gets a chance to listen seriously, one tends to turn up the volume to a "realistic" level
That leads to a supplementary question: is there a practical, sensible amplifier power minima for typical Harbeth/BBC monitor style speakers? Is it, could it ever be, 7W i.e. a 300B tube amp or similar?
 
Attachments only viewable to members

gakyle

New member
Are we accustomed to clipping, 'most of the time'?

Are we accustomed to clipping, 'most of the time'?

From "fluidity" to "clipping" in ninety not so easy steps. The piano extract sounds as though there is a little compression on the second sample so that is the clipped one to my ear. However, if as MH suggests, 'most amplifiers clip most of the time', how is a non-professional not to become conditioned to the sound of clipped music? If this is the case, the sound of non-clipped music might be different enough for it to suggest it is something that it isn't!

If timbre, on brass for example, changes when clipped, it isn't fidelity, high or otherwise, and my "adequate" 50w + 50w amp in a 13'x12'x7' room may be letting me down. Whilst I don't know the true sound of the original instrument, nor the truthfulness of the recording and the production of the CD, I am not experiencing distortion nearly as gross as that on the end of side of an LP. I'm not too worried.

Time for a brief musical interlude? I do enjoy the suggestions occasionally found on HUG. As an example of "fluidity" I think the Oscar Peterson Trio's "Night Train" Verve 314 521 440-2 might be suitable. Nobody swings like Oscar. If an overblown flute could be thought of as an instrument "clipping", Herbie Mann's "Memphis Underground" Atlantic Records K40038 STEREO has plenty. My copy has the advantage of a goodly amount of end of side distortion too.

I just let my very nice, if modest by some standards, stereo, portray the fluidity of the Trio's timing and the overblown distorted timbre of Mann's flute. Even if there is a technical shortcoming or two, this stuff sounds amazing.
 

Pluto

New member
Review archive

Review archive

And that is the playing field for a large majority of valve amplifiers. Perhaps it is how they sound when clipping that is part of the reason for the claimed warm, smooth, liquid valve sound, which may get reinforced by some pyschoacoustics at play. If so, plenty of valve amp enthusiasts are going feel vindicated.
May I suggest that you review this thread.
 

mhennessy

Member
Getting the facts about clipping headroom

Getting the facts about clipping headroom

Please, let's get a little perspective here. I'm being selectively quoted, and it's being taken out of context. Here's what I said:

But the most significant changes can occur if you play with the clipping behaviour. First, it's important to realise that most amplifiers are clipping most of the time. OK, not for background music, but when one gets a chance to listen seriously, one tends to turn up the volume to a "realistic" level - and we do this by making it uncomfortably loud, then backing it down slightly. It's a sub-concious act - we all do from time to time, myself included!
And in my experience, this is very common indeed. I've seen it time and time again - both at work and pretty much every time I've been to a hi-fi dealer. Likewise when I visit friends with hi-fi systems, as they are naturally keen to "show" their systems. Where I can, I put a 'scope across the output to prove the point.

As I've said, finding the "gross" clipping point by ear (where it does actually sound obviously distorted), and turning it down until that just stops, will leave the amplifier in the clipping region. Unlike pure tone, with speech and music, mild clipping is perceived as colouration, not distortion. It's not like photography, where you can see even the slightest hint of overexposure whether it's a test chart or a real scene. Putting it bluntly, just because an amplifier sounds OK, it doesn't mean it's not clipping!

And the context was this: I was responding to a question in the previous post about how amplifiers could be made to sound different at high levels. I was giving my perspective from the point of view of the amplifier designer, and explaining why an amplifier designer will want to pay attention to clipping behaviour. It is my experience that when people compare amplifiers, they do it at high levels. Often well into clipping.



I hope that Mark will soon put his scope across his loudspeaker terminals and tell us what actual voltages are necessary to play this track at a decent level.
This morning, I had a little time to experiment. I need to think of the best way to document it and demonstrate the colouration that takes place in a way that is accessible to all. I'm thinking that perhaps I could take a recording measured at the loudspeaker terminals which could then be hosted somewhere. That will take a lot of effort.

But in the meantime, here are the numbers that show that a 50 watt amplifier can be persuaded to clip at a level that is not much higher than I'd use for serious, critical listening. Using the Diana Krall track, here are the nominal peak voltages for a number of different levels:

  • Background listening: 1.5V
  • Serious, quiet (suitable for late night listening in a terraced house with kids asleep upstairs): 5V
  • Serious, loud: 25V
  • Clipping point: 30V

As you can see, the "serious, loud" case has practically no headroom (around 1.5dB). Just another 2 presses of the "Volume Up" button, which on my preamp moves in precise 1dB steps, puts the amplifier into clipping. The loudspeakers are around 83dB/watt, suggesting peaks of around 100dBSPL at 1 metre from each. The mean value would be some 12-20dB less, so perhaps around 85dBSPL. Now, I'm not suggesting that I'd listen for long periods of time at that level - nor am I suggesting that any of you might - but in the scenarios I mentioned originally, high levels are the norm. Much higher sound levels are encountered at a live concert or night club. Or in a control room when contemporary music is being mixed!

I have a small room - if my audio room was 20+ foot square with decent acoustic treatment, I'd need rather more than 50 watts to be confident that the amplifier never clipped on peaks at even moderate levels.

Remember, every dB requires around 25% more power. 3dB requires a doubling in power. The relationship between watts and volume is incredibly non-linear.

Why do I care about this issue? Mainly because I've seen people dismiss loudspeakers unfairly because they've not recognised that the amplifier was clipping. That matters to me.

All the best,

Mark
 

A.S.

Administrator
Staff member
Appreciate some clarification

Appreciate some clarification

I'm not sure if readers appreciate it, but post #93 by Mark, is extremely profound indeed. I'll explain more later, but for clarity, Mark, can you please (briefly) answer these points for us:


1. How did you measure the voltages you mention, such as '1.5V' .... with what equipment? Could you give us an annotated photo of that equipment in use, connected across the speaker?

2. Were these voltages truly peak to peak voltages, or some sort of averages ...

3. How did you determine that with a couple of clicks more volume, your amp was working flat out, at its absolute maximum power delivery capability?

4. Do you know at what voltage (3), peak power was delivered?

5. What do you calculate as the true maximum power potential of the amp, at least with these particular speaker loads, as opposed to what the generic sales specification said, both channels driven?

6. How did you convert the speaker rating of 83dB/W into a rating of '100dBSPL at 1m from each'? With both speakers playing, and assuming a central (mono) sound, such as voice, what would you estimate the SPL (centre stage) at 1m, 2m, 3m away from the speakers peak and average?

7. What do you mean by 'mean value 12-20dB less...'?


Thanks you for taking the trouble to move this on. Your responses above will help me towards a simple summary of great application to this amp debate.
 

Pluto

New member
Is there such a thing as too small an amp?

Is there such a thing as too small an amp?

...post #93 by Mark, is extremely profound indeed
The question this begs, for me, is how on earth can a (for example) 20W amplifier be of any use at all unless the listener has 'unnaturally' efficient speakers or confines himself to inordinately quiet listening?

Perhaps the answers are here. Sorry to labour the point, but I think this is very significant.
 

mhennessy

Member
Technical clarification

Technical clarification

1. How did you measure the voltages you mention, such as '1.5V' .... with what equipment? Could you give us an annotated photo of that equipment in use, connected across the speaker?
I used a Philips PM3217, which is a 50MHz oscilloscope. An old one, but a good one! Do you really want a photo? It's just plugged into the loudspeaker terminals. If so, what do you want on the screen - some clipping?


2. Were these voltages truly peak to peak voltages, or some sort of averages ...
These were peak voltages. I recorded peak to be consistent with a result Pluto posted in post #71. They will be "nominal", as I said, perhaps only accurate to a few %, but close enough for an initial investigation. My Fluke CombiScope will capture peaks more accurately, if required.

3. How did you determine that with a couple of clicks more volume, your amp was working flat out, at its absolute maximum power delivery capability?
The peak voltage from the amp with typical programme is around 30 volts. My preamp changes volume in 1dB steps. 25 volts plus 2dB is, as near as makes no difference, 32 volts. So it will clip. Obviously, this is a little "nominal", as it depends on the exact voltage on the power supply capacitors, and this will be programme-dependent.

4. Do you know at what voltage (3), peak power was delivered?
With a peak voltage of 25 volts, you can say that the instantaneous peak power is 78 watts. Or just over 100 watts at clipping. Not bad for a 50 watt amp!

5. What do you calculate as the true maximum power potential of the amp, at least with these particular speaker loads, as opposed to what the generic sales specification said, both channels driven?
There is no sales specification, as I built it :)

From memory, it delivers a fraction over 50 watts per channel into 8 ohms with two channels driven with a 1kHz sine wave to 1% THD+N. It's actually a 4 channel amplifier, but I didn't try all four together because I didn't have enough dummy loads available at the time! When bridged, it's comfortably more that 150 watts per channel - can't remember the exact figure. The mains transformer is 400VA! If it helps, it's no bother to re-measure it...

These numbers are "continuous average sine wave power" - true watts, if you like. What some people might incorrectly describe as "Watts RMS".

6. How did you convert the speaker rating of 83dB/W into a rating of '100dBSPL at 1m from each'? With both speakers playing, and assuming a central (mono) sound, such as voice, what would you estimate the SPL (centre stage) at 1m, 2m, 3m away from the speakers peak and average?
OK, you know better than I that assessing this is difficult in a real room. So, all I could do is use the specified efficiency rating of the loudspeaker. From that, converting watts into dBw is easy (10Log(power)). Trying to guess at how the SPL will vary at different distances in a real room with reflections - and with two loudspeakers - is not something I'd like to do. The obvious answer is to use an SPL meter; I'll try to borrow one from work next week. Easy to repeat the experiment because I made a note of the preamp volume settings.

10 log (50) is 17dBw, so add that to 83 to get 100dBSPL. From one loudspeaker, at one metre. I'm sat ~2.5 metres from each of them. Perhaps you could argue that it should be 10 log (78), which gives 102dBSPL. Of course, the mention of dBSPS levels was only to give an approximate idea of levels in the room - and these are peak levels, not average levels (see next answer).

7. What do you mean by 'mean value 12-20dB less...'?
The above levels were peak levels. When judging loudness, the ear is approximately "RMS sensing". So we need to know the PMR of the music to know how loud we thought it really was - the peaks don't tell us much, if anything about this.

The value of 12-20dB came from earlier conversations about that DK track, based on our observations of the waveform.

In Audacity you can estimate the PMR of a track from watching the level meters, which show both peak and RMS levels. The PMR of this track varies between ~18 and 14dB during that "middle" section of the track. So, we can deduce that the average level 1 metre from a loudspeaker will be in the mid-80s.

Thanks you for taking the trouble to move this on. Your responses above will help me towards a simple summary of great application to this amp debate.
No problem :)

BUT - before we draw any conclusions, we need to discuss my further work. In short, all of this is strongly programme-dependant (no surprise there!). For example, I can't listen to a RHCP track at more than about 10 volts peak, yet I can't play "Money for Nothing" at the level I want without severe clipping. For brevity I'll say no more at this point, but will happily post the initial data when asked.

Hope this helps,

Mark
 

A.S.

Administrator
Staff member
Photos, and more sensitive speakers?

Photos, and more sensitive speakers?

Very good Mark. Thanks for the trouble. Just what I anticipated you would report.

Yes, to set the scene, I'd really like a picture of your digital scope with a memory setting of peak output on sine wave as an overlay, and a live screen shot of the music below and at clipping (not too much zoomed in - perhaps 2 seconds of x axis. If you can make a PostIt pointer showing the max possible output (or point with a pen etc.) then we can make sense of the display. You should now be able to attach it to this tread, 1280 x is ideal.

One more thing for clarification: how different would the overall situation be if the speakers were somewhat more sensitive? The ones you used had 83dB sensitivity like LS3/5a or similar quality mini monitor. If the speakers were a more typical 85/86dB efficiency, how would that change the numbers, and the number of clicks on the amp before clipping do you think?

P.S. Regarding (3) above: I assume that to determine the maximum power output of the amp (actually, the maximum voltage output of the amp) you play a signal into the amp, connect the load or the speakers and turn up the volume, click by click, observing the increasing vertical magnitude of the display on the 'scope connected across the output. The vertical divisions on the scope are known (so many volts per cm), and you keep turning up the volume until there is no further increase in the height of the waveform traced on the 'scope. When you reach thae volume setting where no more output can be extracted from the amp, you read off the height of the displayed waveform in cms, and multiply that by the number or volts per cm displayed, and write down your stated 'absolute maximum output voltage at point of clipping' i.e. the amp is working flat-out. Correct?

I would also anticipate that when the music/test signal takes the amp into clipping, the displayed wave momentarily becomes a brighter, whiter, spot on the screen, making clipping point visually unmistakeable. True? If so, a photo of that would explain the issue.
 

mhennessy

Member
Further analysis of clipping in action

Further analysis of clipping in action

Hi,

Sorry it's taken a while. Extracting the gear from the workshop and setting it up does take a bit of time, and taking photographs does take some trial and error...

Yes, to set the scene, I'd really like a picture of your digital scope with a memory setting of peak output on sine wave as an overlay, and a live screen shot of the music below and at clipping (not too much zoomed in - perhaps 2 seconds of x axis. If you can make a PostIt pointer showing the max possible output (or point with a pen etc.) then we can make sense of the display. You should now be able to attach it to this tread, 1280 x is ideal.
OK, if I've understood correctly, this picture should be what you're after:

You must be registered for see images attach



You can see that there is a small amount of "dynamic headroom", as the average level of programme material is less than a sine wave, so the power supply rails will be slightly higher. With this amp, there isn't much in it - less than a decibel.

I'm impressed at how closely the numbers agree closely with my visual estimates from the analogue 'scope the other day. Of course, this 'scope is measuring peak-to-peak, and I recorded peak, so my earlier results would need to be doubled. I said 25 volts for "loud serious", and 30 volts for "clipping", and we see 51.6 and 66.1 respectively. Not bad...

One more thing for clarification: how different would the overall situation be if the speakers were somewhat more sensitive? The ones you used had 83dB sensitivity like LS3/5a or similar quality mini monitor. If the speakers were a more typical 85/86dB efficiency, how would that change the numbers, and the number of clicks on the amp before clipping do you think?
Now, as I'm sure you know, this is not an easy question!

If somehow we were able to find two loudspeakers that were completely identical apart from sensitivity, then we could say that we'd need less power from the amplifier to achieve the same sound levels. And this is not to be sniffed at, as 2 or 3 dB more sensitivity equates to a power requirement of 31 or 25 watts instead of 50. A reasonable saving in terms of power supply, heat sinking, etc...

But of course it's not that simple. For a start, even just measuring sensitivity is a complex business, so how sure are we that the supplied figures are accurate? And our perception of loudness is very frequency-dependant, so if the less efficient loudspeaker had a more "forward" midrange, say, compared to the more sensitive unit, they might sound very similar to us. Of course, many other effects contribute...

So with all that in mind, I decided to repeat the listening with a loudspeaker with a sensitivity of 85dB/watt. But the subjective results were inconclusive. Perhaps the more efficient model was slightly louder for a given volume setting. But perhaps the slightly smoother midrange of the more efficient model convinced me to listen louder? But trying some different tracks confuses the picture entirely - something bass-heavy like Massive Attack sounded better on the less efficient model, whereas something "dry" like Donald Fagen worked slightly better on the more efficient units. I really wouldn't want to call it. And this is no surprise. Too many variables...

P.S. Regarding (3) above: I assume that to determine the maximum power output of the amp (actually, the maximum voltage output of the amp) you play a signal into the amp, connect the load or the speakers and turn up the volume, click by click, observing the increasing vertical magnitude of the display on the 'scope connected across the output. The vertical divisions on the scope are known (so many volts per cm), and you keep turning up the volume until there is no further increase in the height of the waveform traced on the 'scope. When you reach the volume setting where no more output can be extracted from the amp, you read off the height of the displayed waveform in cms, and multiply that by the number or volts per cm displayed, and write down your stated 'absolute maximum output voltage at point of clipping' i.e. the amp is working flat-out. Correct?
That's one way to measure it, and it's what I did to get the numbers on the above image.

One of the most widely-used methods is to increase the level of a 1kHz sine wave until the measured THD+N reaches 1%, then measure the voltage (and calculate the power from that). When clipping occurs, we suddenly see harmonics of the original 1kHz tone, at 3kHz, 5kHz, 7kHz, etc., and additionally, we'll usually see some 100Hz from the power supply (most amplifiers use unregulated power supplies - and there are compelling reasons to do so - so as the output transistors saturate during clipping, they pass the unfiltered 100Hz straight out to the load). The THD+N test will catch harmonics and/or mains ripple. It works well, and resolving 1% doesn't require state-of-the-art equipment.

You must be registered for see images attach


(Note that the 'scope was in analogue mode for this. When I captured the sine wave in digital mode for the overlay in the first image, for clarity I used "averaging" to remove the "fuzziness" at the peaks).

As I had the dummy loads and test set in the room, I measured my amplifier using this approach, and measured 21.45 volts RMS (using a True-RMS DVM - easier and more accurate than a 'scope). This equates to 57.5 watts. And that was with two channels working flat-out. With a single channel driven, the voltage was 22.4V RMS, which is 62.7 watts. Slightly higher, as one would expect, but not much as it has a decently "stiff" power supply...

The "estimate from the 'scope screen' approach gave 56 watts, so it's nice and close - within experimental error.

I would also anticipate that when the music/test signal takes the amp into clipping, the displayed wave momentarily becomes a brighter, whiter, spot on the screen, making clipping point visually unmistakeable. True? If so, a photo of that would explain the issue.
Yes, very much so - on an analogue 'scope. Be careful with digital 'scopes because many of them display the image quite differently, and every "pixel" is shown at the same brightness. Some, like Tek's "Digital Phosphor" series attempt to simulate the effect of analogue CRT displays...

But having said that, it's really, really hard to photograph!

You must be registered for see images attach


I hope this helps. I'll leave the gear in the music room for a while longer, so if you need anything more, please shout :)

All the best,

Mark
 
Attachments only viewable to members

A.S.

Administrator
Staff member
Analysis of test CD (Diana Krall)

Analysis of test CD (Diana Krall)

Mark,

I have received in the post this morning a copy of Diana Krall's from this moment on purchased on line (HMV retail store didn't have it). Catalogue B0007323-02. I have no idea if this disc is genuine or not - it looks perfectly normal. I purchased it to repeat your measurements.

Unfortunately, regarding track 5 (the title track), a quick visual examination of the waveform on this CD shows that from the first bar, the audio is markedly level limited and asymmetrically so, not that asymmetry is unusual with musical instruments or voice. I attach a few screen shots with Audition set-up as I normally use it, with 0dB, peak level, full modulation occupying the entire vertical scale. Left channel on top of the right channel. I have used two different ripping tools with identical results.

I suppose that it could be argued that the audio is not actually clipped (although it looks like it is) but severely limited. It has clearly been adjusted during production/mastering with intent to be banging hard against the end stops right throughout this track. We can use software tools to reduce the loudness and let the software predict (by maths) how much the clipping/limiting has removed, and an attempt can be made to restore what has gone.

Can you compare this with your disc please?

The essence of the work you have undertaken in this thread is to illustrate how the amplifier is readily driven into clipping by real-world music. That surely mandates that whatever source audio is used to illustrate that point (especially to a non technical audience) must be beyond technical reproach. If the source audio is itself level maximised (i.e. clipped or seemingly clipped) then that has the potential for great confusion. It also makes it more difficult (or impossible) to demonstrate in isolation the amplifier's clipping action alone.

>
 
Attachments only viewable to members

mhennessy

Member
Limited yes, but of value

Limited yes, but of value

Hi Alan,

Can't immediately find a catalogue number, but the number on the bar-code is 602517050426.

No matter; from a close examination of the opening bars, it looks identical to mine.

Yes, it's not clipped on the CD - else Audacity would show red lines over the waveform - but there's no doubt that some limiting was expertly applied during mixing and/or mastering, and that occasionally looks like clipping in the editor (although on an analogue 'scope, it does look slightly different to amplifier clipping). It's a busy track, especially compared with the rest of the album, and it certainly makes an impact!

And I certainly agree that we should be using material that should be more "blameless" in this regard. This particular track only entered the discussion because it's the track that first convincingly opened my ears to this particular issue.

It's a shame that we only have access to the published CD. But, as I said earlier, comparing the released version to a version that has been clipped by an amplifier does yield audible differences, so it's a valid - albeit non-perfect - demonstration. A good starting point.

For me, there are two distinct threads to this investigation:

1. Clipping happens more often that we might expect - especially with well-recorded material.
2. When it occurs, it isn't always noticeable as obvious distortion, but it can change the character of certain instruments.

Earlier I mentioned "Money For Nothing" - the original 1980s release has an astonishing Peak to Mean Ratio for a contemporary release, and I'm pretty sure that there is no clipping inherent in the recording. Trying to play that at reasonable volume via my 50 watt setup definitely resulted in very visible clipping - and it was audible too. But I'm not convinced that a snare drum - no matter how well recorded - will help to demonstrate point 2 here :)

I think your knowledge of classic music is wider than mine - my collection is roughly 5:1 in favour of non-classical works - so I'm thinking that having identified brass as a potentially good way to demonstrate this effect, there must be some examples of brass-rich music from the classical repertoire that would have been recorded in a more sympathetic manner, and these might be a more appropriate choice for further investigations. I'll have a think.

Any suggestions? I welcome the chance to broaden my CD collection!

All the best,

Mark
 
Status
Not open for further replies.
Top