Amplifier peak power, class D vs class A/B

[scrubb]

I've been looking at amplifiers and I have noticed that Class A/B amps typically have a peak power output much greater than max power. Class D amps however, don't have any peak power. What are the practical ramifications of this difference? Will it make a difference in how an amp will perform or sound? It seems to me you might need a more powerful class D amp to achieve the same performance as a less powerful class A/B amp.

Here's from the review of the NAD 2200:

"We can see a kink in distortion when we hit 200 watts as the unit sails past that to produce whopping 337 watts per channel, both driven! Per design characteristics, you can have much more during momentary peaks:"

And this is from the review of the Buckeye Hypex NC252MP:


"Typical of these class D amplifiers with regulated power supplies, there is no peak power:"

 

[solderdude]

The 'whimpier' the power supply (the more it sags under load) a non regulated power supply is the higher the peak power can be.

In this case, when both amps are spec'd similarly at continuous power ratings it means there is 2dB more headroom in the amp with the non regulated power supply. Maybe even a tad more.

With tube amps there is second effect as well. When a tube amp and SS amp are both spec'd 30W for instance the tube amp can sound louder.
The reason for this is that tube amps, unlike SS amps, deliver power where SS amps deliver a voltage. When the speaker impedance is much higher in the lows (and most energy in recordings is right there) then the output voltage at lower frequencies can exceed that of what a similar rated SS amp would give and thus play audibly louder.

 

[Kachda]

The 'whimpier' the power supply (the more it sags under load) a non regulated power supply is the higher the peak power can be.

In this case, when both amps are spec'd similarly at continuous power ratings it means there is 2dB more headroom in the amp with the non regulated power supply. Maybe even a tad more.

With tube amps there is second effect as well. When a tube amp and SS amp are both spec'd 30W for instance the tube amp can sound louder.
The reason for this is that tube amps, unlike SS amps, deliver power where SS amps deliver a voltage. When the speaker impedance is much higher in the lows (and most energy in recordings is right there) then the output voltage at lower frequencies can exceed that of what a similar rated SS amp would give and thus play audibly louder.

Sorry what do you mean ? P = VI = V^2/R. so for a given speaker impedance, 30W is 30W and the voltage/current depends on the impedance correct ?

 

[sergeauckland]

Some years ago, there was some research done into this, but now I can't remember the source.
Amplifiers were compared with conventional (sagging) supplies and stabilised supplies, as follows:-

1) An amplifier of X watts continuous, but with a sagging supply, giving a dynamic (music) power rating of + Y watts.
2) An amplifier of X watts continuous, but with a stabilised supply, giving a dynamic (music) power rating = to the continuous rating
3) An amplifier of X+Y watts continuous, but with a stabilised supply, giving a dynamic (music) power rating of X+Y watts

What was found in blind tests is that amplifier 1), the one with the sagging supply, was preferred to Amplifier 2), the one with the stabilised supply when auditioned with music.

However, amplifier 3 was preferred to either amplifiers 1 and 2.

The conclusion was that firstly, higher amplifier powers were preferred, given that it was less likely that amplifier 2 was clipping, but amplifier 3 would similarly avoid clipping, but have a more stable output due to the power supply not sagging and the hum content not increasing.

There's one glaring thing wrong with this experiment, that Amplifier 1 may have been allowed to clip (this wasn't specified in the report I read), but the idea that avoiding clipping by having a sagging supply that provides more dynamic power (music power) over a stabilised supply of equal continuous power results in better sound, I think is clear.

Class D amplifiers are usually driven by a SMPS, where the output voltage is stabilised, such that the amp doubles its power when the load impedance halves, and as such will be somewhat like Amplifier 2.

S.

 

[solderdude]

Sorry what do you mean ? P = VI = V^2/R. so for a given speaker impedance, 30W is 30W and the voltage/current depends on the impedance correct ?

For tube amps things aren't really simple.

An amp that is specified as 30W into 4 Ohms @1kHz = 11V

Let's assume amp X has a stabilized supply (SS amp), amp Y has a 'normal' supply (also SS amp) and amp Z is a tube amp with an output transformer.

Amp X will have a continuous and max output voltage of 11V. regardless if the load is 4 Ohm (30W), 8 Ohm(15W), 16 Ohm(7.5W) or even 25 Ohm(4.8W)

Amp Y will have a continuous output voltage of 11V in 4 Ohm (30W) but may be 11.5V in 8 Ohm (16.5W) and maybe even 12V at 25 Ohm(5.8W)
On top of that the peak power may be 13V in 4 Ohm (42W) but may be 13.5V in 8 Ohm (22.8W) and maybe even 14V at 25 Ohm(7.8W)

For tube amps, amp Z, this (depends on the actual circuit) is even more extreme compared to amp Y and when the impedance of the speakers rises the actual output voltage at that frequency could well be 16V instead of 11V at those loads and thus have much more headroom than SS amps.
Because of this (and low frequencies being highest in music signals) a tube amp with 30W nominal rating may well sound 3 or more dB louder while having the same continuous rating it could well sound equally loud as say a 60W amp in SS before clipping levels are reached.

 

[sergeauckland]

For tube amps, amp Z, this (depends on the actual circuit) is even more extreme compared to amp Y and when the impedance of the speakers rises the actual output voltage at that frequency could well be 16V instead of 11V at those loads and thus have much more headroom than SS amps.
Because of this (and low frequencies being highest in music signals) a tube amp with 30W nominal rating may well sound 3 or more dB louder while having the same continuous rating it could well sound equally loud as say a 60W amp in SS before clipping levels are reached.

Especially as one indicator of subjective loudness is distortion (the more distorted signal sounds louder) so a valve amplifier with inherently more distortion will sound louder than a clean SS amp. This is especially so at high powers where a SS amp will maintain low distortion right up until clipping, whereas a valve amplifier will have distortion progressively increasing as output level increases.

S.

 

[Abcdav]

correct me if I am wrong but the nad 2200 is a class G design not AB design. Peak power, relatively to continous power, is expected to be much higher than conventional class AB design.
NAD 2200 was rated 100 watts continuous and 400 watts peak (+ 6dB)

 

[dfuller]

Okay, so this is down to a regulated vs unregulated power supply. Unregulated linear supplies can have a serious amount of short term current storage (depending on capacitor bank size) that switching supplies (which are by nature regulated) just don't have. An amp can draw significantly more current (and therefore generate significantly more power) with an unregulated supply, if the supply is robust enough, for short transients.

That's pretty much it.

 

[solderdude]

correct me if I am wrong but the nad 2200 is a class G design not AB design.

There are also amplifiers like some of the older NAD (power envelope they called it) where for short moments there is a higher power-rail voltage (internally) available that can only be used for a short moment. (indeed the 2200 is one of those and why it had so much peak power)
This is not a 'regular' non regulated supply but a special circuit.

Older Carver PA amps also had 2 or 3 power rail voltages (internally) that only would do something when they are called upon.
This was a way to have high power capabilities (in class-AB) where not a lot of power was consumed in idle or 'normal' conditions.

 

[b1daly]

Some years ago, there was some research done into this, but now I can't remember the source.
Amplifiers were compared with conventional (sagging) supplies and stabilised supplies, as follows:-

1) An amplifier of X watts continuous, but with a sagging supply, giving a dynamic (music) power rating of + Y watts.
2) An amplifier of X watts continuous, but with a stabilised supply, giving a dynamic (music) power rating = to the continuous rating
3) An amplifier of X+Y watts continuous, but with a stabilised supply, giving a dynamic (music) power rating of X+Y watts

What was found in blind tests is that amplifier 1), the one with the sagging supply, was preferred to Amplifier 2), the one with the stabilised supply when auditioned with music.

However, amplifier 3 was preferred to either amplifiers 1 and 2.

The conclusion was that firstly, higher amplifier powers were preferred, given that it was less likely that amplifier 2 was clipping, but amplifier 3 would similarly avoid clipping, but have a more stable output due to the power supply not sagging and the hum content not increasing.

There's one glaring thing wrong with this experiment, that Amplifier 1 may have been allowed to clip (this wasn't specified in the report I read), but the idea that avoiding clipping by having a sagging supply that provides more dynamic power (music power) over a stabilised supply of equal continuous power results in better sound, I think is clear.

Class D amplifiers are usually driven by a SMPS, where the output voltage is stabilised, such that the amp doubles its power when the load impedance halves, and as such will be somewhat like Amplifier 2.

S.

The major glaring issue with such a test is that driving amps to their limits will make it impossible to level match which makes such a test pointless

 

[bigguyca]

For tube amps things aren't really simple.

An amp that is specified as 30W into 4 Ohms @1kHz = 11V

Let's assume amp X has a stabilized supply (SS amp), amp Y has a 'normal' supply (also SS amp) and amp Z is a tube amp with an output transformer.

Amp X will have a continuous and max output voltage of 11V. regardless if the load is 4 Ohm (30W), 8 Ohm(15W), 16 Ohm(7.5W) or even 25 Ohm(4.8W)

Amp Y will have a continuous output voltage of 11V in 4 Ohm (30W) but may be 11.5V in 8 Ohm (16.5W) and maybe even 12V at 25 Ohm(5.8W)
On top of that the peak power may be 13V in 4 Ohm (42W) but may be 13.5V in 8 Ohm (22.8W) and maybe even 14V at 25 Ohm(7.8W)

For tube amps, amp Z, this (depends on the actual circuit) is even more extreme compared to amp Y and when the impedance of the speakers rises the actual output voltage at that frequency could well be 16V instead of 11V at those loads and thus have much more headroom than SS amps.
Because of this (and low frequencies being highest in music signals) a tube amp with 30W nominal rating may well sound 3 or more dB louder while having the same continuous rating it could well sound equally loud as say a 60W amp in SS before clipping levels are reached.

It should be noted that a peak power test, CEA-2006/490A, allows 1% distortion, that is, -40dB. At average rated power the distortion of a decent amp is .01% to .001% or -80dB to -100dB. The added power for the peaks comes at the expense of 40dB to 60dB more distortion. This is 100 to 1,000 times more distortion for those not into dB's.

The peak measurements are typically done at 1kHz. 20 cycles at peak and then 480 cycles at peak -20dB. The goal of peak measurements appears to be obtaining a high output power value for advertising purposes no matter the level of distortion or the matching of the test conditions to the real world.

 

[solderdude]

The peak measurements are typically done at 1kHz. 20 cycles at peak and then 480 cycles at peak -20dB. The goal of peak measurements appears to be obtaining a high output power value for advertising purposes no matter the level of distortion or the matching of the test conditions to the real world.

I think it is to measure the peak levels closer to what an amp in reality would reach. Yes, distortion will be higher when the internal powerlines sag and 100/120 Hz modulation may find its way into the signal as well.

And then there was the PMPO rating which only has one goal. To create impressive output power numbers for the sake of it.
a 2x6W rated cheap stereo set could be sold as 100W PMPO.

 

[Vasr]

It should be noted that a peak power test, CEA-2006/490A, allows 1% distortion, that is, -40dB. At average rated power the distortion of a decent amp is .01% to .001% or -80dB to -100dB. The added power for the peaks comes at the expense of 40dB to 60dB more distortion. This is 100 to 1,000 times more distortion for those not into dB's.

The peak measurements are typically done at 1kHz. 20 cycles at peak and then 480 cycles at peak -20dB. The goal of peak measurements appears to be obtaining a high output power value for advertising purposes no matter the level of distortion or the matching of the test conditions to the real world.

Generally in content that requires such transient headroom, any increased distortion is not an audible concern. Typically, these additional power requirements, if at all, come in lower frequencies where the ear is even less sensitive to distortion. However, not having sufficient head-room can have audible impact in certain types of content, the characteristics of speakers, how much power you have to start with, etc. It is much more of a concern when you have an under-powered amp to start with like most used to be earlier.

On a secondary note, amps built with good peak power margins correlationally (rather than causally) tended to be built more robustly than those without. But not a perfect correlation.

 

[sergeauckland]

The major glaring issue with such a test is that driving amps to their limits will make it impossible to level match which makes such a test pointless

It's possible to match gain, so for any input, each amplifier gives the same output, up to itslimits. The amplifier audio circuits were identical, the only difference being their power supplies.
The point of these tests was to assess preference, under blind and level matched conditions, with volume uncontrolled, such as it would be in normal home use, and to test for any correlation between such preference and peak music power output, not just continuous power.
I can no longer can remember the details of the amplifier circuit, to be sure that the amplifier design clipped cleanly without latching or other undesirable effects, but assuming that was the case, then the tests seem to me completely valid for what they were trying to establish.

S

 

[b1daly]

It's possible to match gain, so for any input, each amplifier gives the same output, up to itslimits. The amplifier audio circuits were identical, the only difference being their power supplies.
The point of these tests was to assess preference, under blind and level matched conditions, with volume uncontrolled, such as it would be in normal home use, and to test for any correlation between such preference and peak music power output, not just continuous power.
I can no longer can remember the details of the amplifier circuit, to be sure that the amplifier design clipped cleanly without latching or other undesirable effects, but assuming that was the case, then the tests seem to me completely valid for what they were trying to establish.

S

If the amps were level matched with output into a speaker at a lower volume, then the test conducted at the limits of the power amps capability then the comparison might provide some real world relevance. Essentially it would provide guidance about how much rated power you need for your purposes given the type of amp.

Otherwise the louder amp wins in listening comparisons.

 

[restorer-john]

The first successful commercial class 'G' was built by Hitachi. The HMA-8300, long before NAD came out with the Powertracker (2200) or Proton (D1200) with their DPD (dynamic power on demand).

They use a perfectly valid design in my opinion, although as a class, they fail way too often. Switching in HV rails at high power means large amount of collateral damage in silicon when they ultimately die. All the major brands built commutating rail amplifiers in the early to mid 1980s. Pioneer, NAD, Marantz, Hitachi, Kenwood, Yamaha etc. All but NAD abandoned them a few years later and moved to high current designs offering better performance on a continuous basis into the lower impedance speakers that were entering the market in the late 80s early 90s. A lot more nominal 6R and 4R speakers.

A tightly regulated supply and amplifier design is no more "correct" than a supply designed to allow high momentary peaks. As long as the continuous rating is real, the high dynamic performance is a bonus. Where the design is all about the dynamic power, and you end up paying for that over continuous, it's not such a good tradeoff in my opinion. It's the turbo 2 litre vs a 5.7 litre V8 analogy.

As for the EIA being 1% THD for 20mS 0dB/480mS -20dB tonebursts, I have no problem with that either. I've laid out the EIA and EIAJ/JIS dynamic power test details before in ASR threads. They are different to one another, but in typical amplifier tests, the measurements vary little. I use the old EIA one, simply because the oscilloscope/DSO can lock on easier than the EIAJ/JIS one where there is an absense of signal in between the bursts.

 

[restorer-john]

For anyone wanting some high quality sources for their own toneburst testing, I've uploaded four tracks from one of my industry test discs. Details are in the file name. Maximum level is 99.9% of full scale (0dBFS), so no clipping. Download them and enjoy.

1001Hz Toneburst (EIAJ) https://drive.google.com/file/d/1IjKAZAWE7I5yZYs66qZvL_t2swCiNM43/view?usp=sharing
401Hz Toneburst https://drive.google.com/file/d/1Vw5B0rDGM4JHhQ4sD79CD0pJHtzTB_eG/view?usp=sharing
4000Hz Toneburst https://drive.google.com/file/d/1f0VImr8OuPmNtGNONROZIhtBfJm6sibR/view?usp=sharing
1001Hz Toneburst (EIA) https://drive.google.com/file/d/1mbwMXA5Oc7mDXNZ94lWLZMWssMXJR6bw/view?usp=sharing

 

[Phorize]

This is a very interesting discussion.
Just to check if I’m following correctly; where all other things are equal an unregulated supply will lead to less distortion with demanding transients, although the degree to which this is audible will vary according to the degree of power available in the first place and the specific speaker demands, content etc. However the average asr member is unlikely to be impacted as they will be running a high power amplifier. In my case a March audio p452 which easily has a enough power to destroy my harbeth p3s with their 50w load.

I have always wondered about this ever since I bought an early 80’s cd release of this album in the early 90’s:

Tantalisingly it came with a warning sticker on the case that said ‘Contains transients that may damage your system’. I blasted it through my nad 3020/celestion ones and nothing bad happened except the predicted gentle clipping

 

[Armand]

Okay, so this is down to a regulated vs unregulated power supply. Unregulated linear supplies can have a serious amount of short term current storage (depending on capacitor bank size) that switching supplies (which are by nature regulated) just don't have. An amp can draw significantly more current (and therefore generate significantly more power) with an unregulated supply, if the supply is robust enough, for short transients.

That's pretty much it.

It seems most people have an understanding that all SMPS are regulated. That is not correct. The popular Hypex SMPS's are not regulated and output voltage it totally dependant on input voltage just like a regular linear power supply. The "only" difference is a totally different switching frequency and transformer.
With these types of SMPS, big capacitors are just as important as on a linear power supply, but they must be on the high voltage side (just after the rectifier and 325VDC here in Europe). The capacitors on the secondary side can be much smaller because of the higher frequency.

Regulated SMPS'es that try to keep the voltage constant on the output (by using feedback) can introduce modulated noise on the power rails if the feedback control loop is not fast enough (which it typically not is.)
The perfect power supply is of course one that keeps the rail voltages 100% constant regardless of how many amps drawn with a bandwith of at least 20kHz, but that would be very expensive and not really pay off. Especially not on todays class D amplifiers with a lot of feedback that just corrects everything anyway.. On power amplifiers without feedback it is a totally different story.

Anyway, when buying an amplifer you should buy one that is capable of delivering the power you need before it approaches clipping. When operating close to clipping the distortion values are often unpredictable and often not measured properly either.

 

[Impossible]

does the peek power effect the dynamics if e.g.

Class A/B, Max power 300w, Peek power 500w
Class D, Max power 500w, Peek power 500w

If max power is the same, would there be a difference in dynamics?