How much amplifier power do you really need?

[RebelMan]

It puzzles me why you found the 170 so lacking as compared to the 240 when driving the Kef Blades.  I suspect the conditions between listening sessions differed significantly.

The amps were swapped in front of me, with everything else left unchanged, the same music was used too.  The D170 went first, and had been powered up and running for some time, so arguably it had an advantage over the "swapped in" D240.

Someone once said if you eliminate the impossible whatever remains however improbable must be the truth.  The 170 and 240 are bred from a similar (nearly identical) architecture.  Applying the aforementioned rule it would be impossible for the 170 to sound anything short of similar to the 240.  All that remains is the conditions of the test environment.  Nothing else changed?  How can you be so sure?  How did you confirm everything?  Surely you took level matched measurments before and after the swap?  Surely you measured the SPLs at the same position using the same sources before and after the swap?  See where this is going?

[RebelMan]

@RebelMan: To be frank, a lot of what you wrote above doesn't make much sense to me.  For example:

* What is the distinction you're drawing between "clean" and/or "linear" power and any other kind?  Power is a well-defined, single, measurable quantity so in itself it can't be clean, dirty, linear or non-linear.  Maybe you are talking about distortion instead.

* Why does it matter "at what load" the power was measured?  Power is power.

Your argument based on the loudness of an 85 dB/W (at 1m?)  speaker driven by a continous 350 W (RMS) would be valid.  But that is not the point: the video is talking about instantaneous power needed to drive the speakers with a faithful signal at music peaks.  @Confused helpfully made that clear by quoting the description of the power meters from CH Precision's documentation.

What this discussion revolves around is the crest factor of the music being reproduced.  For some uncompressed music this can be over 20 dB, meaning that to reproduce the full dynamic range of the music the amplifier needs to have more than 100 times as much power available (for peaks) than the average (RMS) value for a given volume.  This is a good article that goes through the maths.

I'll start with the second question first.  Power is not power.  Power is the measure to do WORK or the potential to do WORK.  In the strictest sense (as it applies to electronics) power is the amount of work that is needed to move electrons through a given medium.  When fewer obstacles are present electrons are more free to move and thus more work is needed to keep them moving.  The converse is also true, when more obstacles are present electrons are less free to move and therefore less work is needed to keep them moving.  So unless the system requires large amounts of power less will do.  How can anyone tell how much power they need unless they measured it?

GOOD, power that is sufficient in supply at any given time for any given load.

CLEAN, power that is free from artifacts inherent of the architecture's design and components.

LINEAR, as the demands for power increase (instantly or continuously) the supplies of power also increase in lock step.

The problem with the video is that it does not indicate what the conditions are at that "instant".  What is the load when the amplifier displays 350W?  My example amplifies the problem when snapshots of power are taken out of context.  I made the assumption that the load was 6 ohms the instant 350W was needed to prove a point.

How can you conclude that an amplifier needs 100 times more power available for peaks when we don't even know what those peaks are?  That's purely conjecture.  You cannot solve what you do not know.  Again how do you know?  You need to measure.

[RebelMan]

They need GOOD, CLEAN and LINEAR power and proper testing proves this time and again.

Interesting, can you provide a link to some good examples of such testing.  It is good to add proper test data to the discussion.

I will defer you to John Atkinson and his work illustrating the good the bad and the ugly amplifiers he has tested through painstaking and time consuming analyses.

[RebelMan]

Interesting post, @Confused.  It agrees nicely with what Roger Sanders says in his white paper (which I know I've linked to before).

It appears Roger is making some pretty big assumptions.

Which assumptions in particular do you think invalidate his analysis?

For one, he is assuming... "audiophiles usually are using underpowered amplifiers and are therefore listening to clipping amplifiers most of the time". Which audiophiles?

For another, he is assuming... "conventional, direct-radiator (not horn-loaded), magnetic speaker systems of around 90 dB sensitivity, require around 500 watts/channel to avoid clipping"  Which speakers?

I don't claim that he is invalidating his analysis with the assumptions he is making but he is making them in order to produce his analyses.

Once again he, like other marketing geniuses, are taking things out of context.  At what point is 500W needed?  He doesn’t specifically say.  He is also pro solid-state so it stands to reason he would be biased in his remarks.  Although I am too, he is making a bigger case against SET amplifiers not other quality solid-state amplifiers making his point less considerable in general.

As I read it, his analysis about power requirement is based on straightforward, easily-verified measurements and does not depend on whether the amplifier is solid state or valve.  Calling Sanders a "marketing genius" is clearly intended to be derogatory, and does your argument no favours.

When I checked it seemed pretty evident to me that the man was peddling his wares.  You don't think that some bias was a motivating factor in his discussion? 

I suppose if I needed a little more confidence in my system I would lean on this white paper too.

What exactly are you referring to?

You seem to put a lot of stock in this white paper, especially if you have made reference to more than one time.  It's pretty easy to say "yeah we need more power and this white paper proves it".  Really?  How about trying this one on for size...  If you are so willing to stand by your guy then do what he says.  That's right, do it!  Get an oscilloscope and measure your system just as he describes AND I'll do it too!  Then come back here and describe your system, the room your system is in, how you listen to music (where you sit, how loud you listen, your choice of music) and report what you learn.  I am more than willing to work with you to establish some working parameters that we can both agree on and share.  Are you up for it?

[RebelMan]

I believe I will see the Harbeth's truly shine with the D1000 and look forward to that day.  

Your Harberth’s will truly shine with any Devialet (if power is your concern).  The D1000 is mostly for bragging rights and/or lost hearing.  So be careful playing at those high levels for very long.  At 95dB hearing damage can occur in 60 minutes and remember it’s cumulative.

[Jim_Anderson]

I believe I will see the Harbeth's truly shine with the D1000 and look forward to that day.  

Your Harberth’s will truly shine with any Devialet (if power is your concern).  The D1000 is mostly for bragging rights and/or lost hearing.  So be careful playing at those high levels for very long.  At 95dB hearing damage can occur in 60 minutes and remember it’s cumulative.

LOL

Thank you RebelMan,  

I'm 68 years old and survived the 70's. I do know what I've gained in audio quality as I've moved up through Devialet offerings.  

Have I lost some of my hearing - most definitely BUT ... I can still hear a beautiful women moan, wind whispering through trees when I off road camping, hear trout surface, take a fly and burp.   

Do I play a lot of music - most definitely.  Do I have an excellent library of vinyl and cd's - most definitely.  

When I play a wonderful piece of music and it gets the feet a tapping, I have no hesitation is cranking to 11.  My current configuration and the Harbeths 40.1's are truly a wonderful combination.  On any of the Devialet platforms - Harbeth's perform exceptionally.  I am sure Harbeth owners would agree Devialet/Harbeth are a good marriage.  

What I've gotten in moving up in the horse power range is more clarity - not limited to but predominantly with the lower registers.  When the dial is turned to the left, the music isn't getting louder - its bringing out hidden gems in harmonics, voicing, transients. 

So Mr RebelMan, as for the Devialet D1000 - you and I will have to disagree on it being a bragging rights thing.  I am very fortunate to have my current system. Is it the best - hell no, but it makes this old soul happy and for that I'm quite thankful.

Regards, Jim

[David A]

I've got a different take on peak levels and how much power you need.

Back in 2001 I bought a pair of Dynaudio Contour 1.3 Sys with a rated sensitivity of around 85 dB and I seem to remember seeing a test report that put their sensitivity lower, around 83 dB. One thing I noticed about Dynaudio speakers in general is that it didn't seem to matter how big an amp you had, they always seemed to sound better with a bigger amp. I listen at what I think most people would regard as low levels. When I measured volume levels at my listening position around 2 metres from the speakers most of the time they were running below 80 dB, often close to 70 dB but using the hold function on the meter I was seeing peaks around the 85 dBC mark. The meter could also display instantaneous peak levels and they tended to run in the mid 90's so an amp capable of delivering 100 W into 4 Ohms was certainly capable of delivering the instantaneous peaks I was getting at my listening levels and I never noticed distortion or clipping on peaks. My amp could deliver 170 W into 4 Ohms but despite that I noticed an improvement if I used a more powerful amp. In the end I came to believe that the difference had little or nothing to do with output capability and everything to do with the amp's power supply and it's capability to deliver high power quickly when transients occurred and also it's ability to quickly recover from those high output demands and be ready to keep delivering on demand. Smaller amps tend to have less capable power supplies than larger amps and I ended up deciding that what the bigger amps bought to the table in terms of delivering better sound quality wasn't more output power but rather a power supply that was capable of delivering peak output more capably at your chosen listening level.

When I replaced that 170 W amp in 2016, I replaced it with a Devialet 130 Pro which had less rated output than my previous amp. Despite that, it actually delivered better sound quality and better peaks and transient response than the previous amp. Since sound quality improved with lower peak output capability it was obvious that I wasn't getting clipping with the previous amp, if I had been running into clipping with it then I should have been running into clipping more often with the Devialet but just as I hadn't noticed clipping with the previous amp I wasn't noticing it with the 130. I used SAM with the Dynaudio and didn't notice any issues from a lack of power when using SAM.

Subsequently I replaced the Dynaudio with my current Focal Sopra 2s which have a 91 dB sensitivity rating. Still no noticeable problems. My listening levels haven't changed and I usually set my volume setting at around -20 dB for most of the music I play though with some softer recordings I do increase the volume setting a bit, rarely to more than -10 dB. At those volume levels I could raise the volume setting by over 20 dB, even by close to 30 dB, before I hit maximum volume. My feeling is that for my listening levels the 140 has more than sufficient power and a power supply more than capable of keeping up with the transient demands of the music I listen to which is mostly small group acoustic jazz. Things might well be different if I were playing rock at high levels but that isn't my scene. When I held a meeting of my local audio club here a bit under 2 years ago there were a number of members who do like rock at high levels and they proceeded to "kick the tyres" on my system by listening to the kind of music they like at the kind of levels they like (I left the room and left them to it) and they all seemed to like the result so I don't think the 140 failed to deliver at higher levels with the Sopras.

I do think bigger amps can deliver benefits with many speakers and those benefits may sometimes be related to avoiding clipping at peak levels but I do think more often the benefits of a larger amp are simply related to a bigger and better power supply rather than the ability to deliver more watts without clipping. If I were still listening to the kind of music I listened to in the '70s and 80's at the level I listened at then and were still using lower sensitivity speakers like Dynaudio and/or I had a bigger room then I would definitely be running a bigger amp than the 140 but I'm not and so far the 140 is meeting all of my needs and delivering the best sound quality I've ever had. YMMV.

[thumb5]

@RebelMan

The points you quoted from Roger Sanders' white paper are his conclusions, not assumptions.

They follow pretty much from this simple maths:

Let's say you need around 5 W RMS to achieve normal listening levels (whatever they are) with a low-ish efficiency speaker in a normal-sized room. Given that -- from unbiased sources -- we know some music can have a crest factor of over 20 dB, that means the amplifier has to deliver over 500 W peak power to reproduce that music without clipping.

That seems very straightforward and uncontroversial to me. Which of the two premises do you disagree with, or what is the fault in the logic?

[Antoine]

Just to note; in case of our Devialets the sheer number of variables is even larger and makes this discussion even more difficult as their internal microprocessors in the power supply and output stages run code determining the voltage and current limits short and long term. It’s certainly possible these code imposed limits (said to be there for thermal management and accounting for the differing heat dissipation capabilities of the different cases) could affect SQ e.g. when comparing a 170 vs. 240 or any model. The smaller models like the 130 have a different class D output stage.

[David A]

@thumb5

Your maths are correct in that if your normal listening level requires 5 W and you want to produce peaks that are 20 dB louder then you need to be able to deliver at least 500 W.

The premises/assumptions in your reasoning that I disagree with:

1- that most people needs 5 W for "normal listening level", even with 85 dB sensitive speakers. My normal listening level is around the 80 dB mark. My listening distance is a bit over 2 metres and an 85 dB sensitive speaker would produce roughly 80 dB levels for 1 W output at my listening distance but I do listen in stereo and that means that 2 speakers each producing around 80 dB for 1 W output will deliver an 83 dB listening level at my listening position. We get a 3 dB gain in efficiency from having 2 speakers so I actually need a little less than 1 W to produce an 80 dB listening level at my listening position. OK, my preferred listening levels are on the low side for many people but I do think that 5 W is a little on the high side even for low sensitivity speakers given that we're using 2 speakers to produce that listening level.

2- In order to produce a 20 dB higher peak than our normal listening level we do need to multiply the output required to produce the normal listening level by a factor of 100. Whether or not the music a given person listens to is going to have a crest factor of 20 dB is arguable. It's definitely possible with classical music and other well mastered acoustic music but I think the crest factor for a lot of pop/rock music with limited dynamic range is likely to be significantly less than 20 dB. Depending on your taste in music you may not need to accommodate a crest factor of 20 dB.

3- Even accepting your 5 W starting point for normal listening level and the 500 W needed to faithfully deliver a crest factor of 20 dB, that doesn't mean you need an amp capable of delivering a continuous 500 W output and amplifier power ratings are for continuous output. Amplifiers are capable of delivering more than their continuous output rating for transients without clipping. I've got no idea what output my 140 is capable of delivering on a transient but my normal listening level would require less than 1 W per channel, a 20 dB crest factor would require it to deliver transient peaks of 100 W a channel, and the 140 is rated at 140 W continuous output per channel into 6 Ohms and my speakers are 4 Ohms. That means it's more than capable of delivering a 20 dB crest factor without clipping given my normal listening levels, and it could probably meet that 20 dB crest factor requirement for someone who needed twice as much power per channel for their normal listening level.

Given that we use 2 speakers, that most speakers aren't low sensitivity, and that we don't need an amp capable of continuous output at the level required for instantaneous peaks, I wonder just how real world your 500 W example is. My feeling is that the 140 Pro is sufficient for quite a few people and that the 220 and 250 are adequate for most people who need more than the 140. Some people will definitely require a dual amp setup and there are certainly speakers which place more demanding loads on an amp and benefit from larger amps than these figures suggest plus I think there are some sound quality gains from the dual amps that some people find attractive. I think real world maths suggest that we don't need an amp capable of 400-500 W/ch but there are certainly some people who will need that given their choice of speaker, listening level, and room size and there are others who go for the dual amp setups for reasons unrelated to power requirements.