How much amplifier power do you really need?

[RebelMan]

All this technical stuff above goes way over my head but just below that sit my ears. And they tell me more power means better sound. So can anyone translate the above so my ears understand please?

Yes. Your ears lied!  

It must be the two instead of one power cords that you use with dual mono that makes the difference then. My ears don’t lie. It’s like listening to my dog barking versus the labrador next door. It’s different.   Clearly

It was mostly a facetious answer I gave you, remember you asked.  

If you don't think your ears can deceive you then you clearly have not been exposed to the fine art... ahem... science... of psychoacoustics and psychophysics.

[RebelMan]

There is nothing mysterious.

If you calculate power by squaring the power supply rail voltage and dividing by the load impedance, you're calculating peak power.  Amplifier ratings are normally (at any rate, should be) quoted as RMS with a sine wave signal.  The peak power for such a signal is exactly twice the RMS power.

Example: 140 Pro PSU rail is 41 V, so peak power into a 6-ohm load is 41^2/6 = 280 W, vs quoted power 140 W RMS.

Your math agrees with the calculations I posted, in theory.  In practice, however, peak power is not being reached because of hidden code or the protection circuits are kicking in too soon preventing it.  Audio tested the 140 and could only produce 102 W in to 8 ohms and 203 W into 4 ohms before the protection circuits kicked in.  This was also the case with the 220 when HiFi World tested it, "...using 0.2 sec sine wave bursts, power measured 144 Watts (34V) into 8 Ohms and 290 Watts into 4 Ohms.  Into 6 ohms, burst power is 190 Watts, slightly below the 220W figure quoted but not a consequential shortfall."   

Using theses figures in your example the math becomes... 220 Pro PSU rail is 34 V, so peak power into a 6-ohm load is 34^2/6 = 192 W, vs quoted power 220 W RMS.  Which again supports the observation that peak output power is being limited to continuous output power levels.

Do you know were these measurements done properly - like was the SAM turned off? The SAM can limit the power.

Unfortunately, it is unclear what status SAM was in when the 220 was tested by HiFi World. Since continuous output power was reported half that of the published specifications I tend to agree with @Jean-Marie that SAM was enabled.  Accordingly peak power should be double the 192W that was measured.  However, Audio was able to produce continous output power that did match the published specifications but they were unable to exceed them which leads me to conclude that peak output power was limited to continuous output power levels due to the code and/or the protection circuits.  I't just a guess but one that the facts support.

[RebelMan]

I did a bit of experimentation with the configurator on the D250 which is an interesting case:

For a D250 stereo, the ADH sections is:
"ADH_CONFIG":{
    "VOLTAGE_0DB":55,
    "PUISSANCE_0DB":250
  },

And for the D250 stereo, the same section becomes:
"ADH_CONFIG":{
    "VOLTAGE_0DB":55,
    "PUISSANCE_0DB":1000

55V under 6 ohms is a power of 504W, so about twice the rated 250W RMS

The way you bridge your amplifier (turning it into mono) is that instead of connecting your speaker between the black post (connecting to the ground) and the red post (swinging between -55V and +55V) is that you connect it between the two red posts and you invert one channel. So each red post still swing between -55V and +55V, but because they are in opposite phase, the difference between the two is not double, swinging between -110V and +110V between the two.
Guess what, 110V is 2016W under 6 ohms, giving once again the rated RMS power of 1000W for a pure sinusoidal tone.

@RebelMan what does continuous power means? I am reading as RMS power on a pure sinusoidal tone (or tones) that the amplifier can provide on a continuous basis. With that meaning, the peak power of such signal would still be twice the RMS power. This is intrinsic to the nature of the Sinus.
The only way for it not to be true would be to use a very different kind of signal where the crest factor would be much lower than 3dB, which is possible but quite unlikely in my mind because it would be very different from any measurement protocol of audio equipment that I know of (but I'm not pretending to know every one of them.)

I would like to think that peak power could be reached for a short time too but how short?  According to HiFi World it would have to be less than 0.5 sec.  Fortunately, they were able to mimic peak power occurrences within 0.2 sec.

Is there a typo in the second code snippet?  Shouldn't this...

And for the D250 stereo, the same section becomes:
"ADH_CONFIG":{
    "VOLTAGE_0DB":55,
    "PUISSANCE_0DB":250

read like this...

And for the D250 dual-mono, the same section becomes:
"ADH_CONFIG":{
    "VOLTAGE_0DB":55,
    "PUISSANCE_0DB":1000

While your math computes and your logic is rational I do not think it is applicable in this case because the 250 is not operating in stereo mode when power output is 1000W.

It is my understanding the Devialet Expert 1000 Pro (Dual-Mono) is a bridged mono mode amplifier.  By definition a bridged mono mode amplifier quadruples power output into a given load AND the load doubles.  When the load doubles, the current doubles.  This is why a bridged mono amplifier "sees" half of the impedance.  In other words, an 6 ohm speaker behaves like a 3 ohm speaker to the amplifier.  The math in this case would be computed as follows...

The 250 in stereo mode:  P = (V*V)/Z. P = 250watts and Z = 6ohms therefore voltage V = sqrt (P*R) = sqrt (250*6) = 38.72V. (If VOLTAGE @0DB is 55V then how is PUISSANCE @0DB is 250W interpreted?  Is it a voltage limiter?)

The 250 in bridged mode (1000 dual mono):  Voltage doubles and therefore V = 2*38.72 = 77.44V. P = (V*V)/Z = (77.44*77.44)/6 = 1000W. (1000W does not equal 2016W)

As can be seen by these calculations power quadruples from 250W @6ohms to 1000W @6ohms (as expected) and takes into account the doubling of voltage (38.72 to 77.44 vs 55 to 110 by your figures).  

That said, peak power for a sinusoid is computed as 1.414 RMS power.  Since peak power is 1.414 RMS power, peak power becomes P(peak) = 1.414 * 1000 = 1414W not 2016W.  How did you determine peak power to be twice RMS power?

Since the amplifier sees double the load or half the impedance when in bridged mono mode a 6 ohm load would behave like a 3 ohm load.  If the voltage were held constant on the supply side of the amp at 38.72V (which the "VOLTAGE_0DB":55,"PUISSANCE_0DB":1000 statements seem to imply as 55 is a constant in both code snippets) current would need to double.  To remain stable a bridged mono amplifier would need to see twice the impedance.  Hence, if the amplifier was stable down to 1 ohm in stereo mode then it will only be stable down to 2 ohms in dual-mono mode.  This could explain why HiFi News reported the power output of the Devialet O d’A (for which the 1000 is pattern after) slipped by more than half from 1500W@2ohms to 670W@1ohm. 1500W@2ohms make look impressive compared to the 420W@2ohms that the 140 can spare but it still only amounts to a 6dB gain overall.

[thumb5]

As I pointed out before, for a sine wave the peak power is exactly twice the RMS power, because the (peak) amplitude is the square root of 2 times the RMS value of the signal and power depends on amplitude squared.

A lot of heavy weather seems to be being made of this...

[David A]

This thread has been rolling around in my head for some days now as I tried to make sense of some of the comments plus my own experience with amps which, as was pointed out above, related to class AB amps which behave differently. In the end 4 things seemed to stick with me:

1- @ "Jean-Marie" 's comment that we can know the peak output of a Devialet and his quote from the ADH section of the config file showing the settings for Voltage and Puissance at 0 dB;

2- @RebelMan 's comment that Devialet's have no dynamic headroom,

3- @thumb5 's comments in various places about crest factor.

4- I seem to remember reading a post in which one of the Devialet team was quoted as having said that his perfect amplifier would have infinite power. I haven't been able to find the quote on a quick and rather cursory search so I'm sorry I can't give attributions to the poster or the Devialet team member.

Anyway, what ended up coming out of all of those things rolling around in my head was the following and it may sound surprising. It may also be wrong but I think that logically it does hold together and is consistent with all 4 of the things I listed above. Feel free to call me an idiot if you wish but give an explanation of where you think I'm going wrong.

My conclusions are this:

- We can ignore crest factor. The reason for that is that we're working with a digital device, peak output occurs at 0 dB and the highest level for a digital signal is also 0 dB. Yes, crest factor exists in music but the signal at the peak of the crest can't be encoded at a higher level than 0 dB.

- When we set the volume level we tend to do it based on how loud the music sounds to us at that moment but what we're really doing is setting the output level for a signal strength of 0 dB. If we set volume to -20 dB, the maximum peak in the music will occur 20 dB lower than the amp's maximum output. Since the peak signal level is O db, setting the volume to 0 dB will result in the amp delivering peak output for the highest signal level which can be encoded. We don't need dynamic headroom to handle peaks, the amplifier is basically designed to be capable of peak output continuously which is why it has no dynamic headroom. If things aren't loud enough for us with musical peaks of 0 dB level at a volume setting of 0 dB we need a bigger amp.

- Provided we don't exceed a volume setting of 0 dB, the amp shouldn't clip, at least in theory. In practice I guess it may because things like speakers which present a different load to the amp and things like SAM may make a bit of a difference and result in the amp "running out of steam" a bit below a volume setting of 0 dB.

- Why have a volume control which can set levels above 0 dB if the above is true. Some musical content is mastered with its peak level lower than 0 dB. If the peak level is mastered at -5 dB, you would need a volume setting of +5 dB to reach peak output of the amp and people should be able to reach peak output. Some test content is mastered at levels like -20 dB and people using that test material may need to play that content at the amp's peak output level for their test. There's some good reasons for having a volume control which can go above 0 dB but you would only want to go above a volume setting of 0 dB for signals mastered with a peak level lower than 0 dB

- If the ideal amp is one with infinite power, then an ideal amp will have no dynamic headroom because it will always be capable of delivering as much power as needed for as long as it's needed so it will be capable of infinite power continuously if required and, just based on the meaning of infinite, we can't have "more than infinite" so dynamic headroom for an amp which has infinite power simply doesn't exist, dynamic headroom will never be required.

The Devialet designers seem to have taken that idea of an amp with infinite power and delivered amps which essentially behave the way such an amp would behave but with a limited power output. We get the amp's rated power for as long as we need it but we can't get any more than it is capable of. Peak output is peak output, that defines the level that musical transients (the top of the crest) can reach, and we can stop worrying about crest factor and dynamic headroom provided we don't exceed peak power output. A volume setting of 0 dB is the best guide we have to when that is likely to occur but if you're using SAM and/or have a speaker which can present a difficult load you may want to keep your volume setting a few dB below 0 dB.

That's what I ended up coming to for my conclusions and I have to admit they surprised me. I ended up thinking about the amp's power delivery capability in a very different way to how I thought about such things before this thread. Like I said, i could be wrong.

[thumb5]

@David A At first glance, that looks pretty convincing and a nice resolution of the various strands -- at least, in the context of Devialet amps. Need to give it some more time to sink in or otherwise, but I like your thinking.

[Confused]

@David A see @IanG-UK 's (aka "The General") post per the link below for a mention of the "infinite power" quote. If you have a little spare time, I would recommend reading through the entire thread, between the normal fluff and nonsense there are some very interesting gems of information to be found. I was at the 2016 Oxford Audio event and can confirm that Ian's words are an accurate reflection of what was said. (I recall he sat there making notes in his note book, later asking questions, he's very diligent our Ian)

https://devialetchat.com/Thread-Devialet...-13th-July

[David A]

@Confused

I have some memories of reading that thread back in 2016, that was a few months after I bought my 130 (now a 140). Back then a lot of that discussion went over my head because I was so new to the Devialet range. I certainly have no memory of Ian's mention of "infinite power" in that thread and I have a definite feeling that I saw that comment quoted in a much more recent post somewhere here, but perhaps not in this thread.

I've read Ian's write-ups on the Oxford Audio events each year since then with increasing interest and awareness and if I had to make one comment about his write-ups I'd say that it's a real shame that we seem to get more info from Ian after these conversations with Devialet people than we do from Devialet themselves in the months between each event. Ian has been doing a wonderful job each year at these events and much of what he has been doing is a role I think we should have been seeing Devialet do for themselves. I've a lot to be grateful to Ian for his diligence over the years since I bought my 130 in 2016.

[thumb5]

@David A Maybe you're thinking of @Confused's original post in this thread?

It also reminds me of the day at Oxford Audio, when Mathieu Pernot stated that his perfect amplifier design would have "infinite power".  This seamed slightly absurd to me at the time and I was talking to him about this later in the day.

[thumb5]

@David A's comment in an earlier post got me doing some background reading:

It would be interesting to know how crest factor relates to the dynamic range measurement Roon uses. From reading what Roon have written about the dynamic range measurement in their user guide and other documents, it defines the range in which something like 95% of the music occurs, excluding the loudest peaks and the softest passages. It's not a measure of the range from silence to maximum peak level of the recording.

While not touching specifically on what Roon uses as an indication of dynamic range, this article from Sound on Sound has a lot of interesting observations on measures of loudness, dynamic range and crest factor, in the context of looking at the effects of the so-called "loudness war".  It's quite technical but a good read for anyone so inclined.  Incidentally it includes a range of measured crest factors for a body of 4500 music tracks which more or less agrees with the figures mentioned earlier in the thread.  (Notwithstanding @David A's observation in his recent post, this is still relevant to the topic in general.)