Class D vs Class A or A/B

[B-string]

I do still have my old Yamaha 2700 waiting as a PA backup.........

 

[tombowlus]

A lot of technical talk here going right over my head but I do have a question regarding all this. What, if any, point is there to non class d amps going forward? Take the GK 1001rb for instance vs. the mb800. What advantage would the bigger heavier rb have over the mb?

I realize they are two different styles of amp but beyond any tonal difference, which I assume could be replicated in a class D rb, and maybe the biamp thing, why would someone buy an rb series over an mb?

An excellent question. I'll start off by saying that the MB800 may have ruined me for ever bringing heavier amps/heads to a gig. That being said, the main differences that we are seeing in our testing comes when we apply the burst power tests. Of course, these results vary based upon the design of a given output section (regardless of topology), based upon the power section of the amp, based upon how that amp designer addressed limiting and clipping, and based upon other factors related to the overall design of the particular amp(s) being tested. I'll also say that the more modern class-D heads which we have tested appear to be improving with regard to how they handle burst power and how they respond when pushed to their limits.

There will certainly be some class-D amps which test "better" - even on the burst tests - than some class-AB amps. But at least with regard to the question above, while the GK MB800 (class-D) was able to hit just over 125% of its rated power in our burst tests, the GK Fusion 550 (class-G) was able to hit over 145% of its rated power in this same burst power test [Note: we were not publishing the results of our burst power tests back in issue #3, which contains our Fusion 550 review, but we have run this unit through the current BGM burst power test and it exceeded 800 watts at 4 ohms.]

In typical gigging use, a player may not notice that difference. But to reiterate my answer the question above, the main difference we see between and amongst the various amp topologies we test is how they handle burst demands. Not saying one topology is better/worse than the other, only that this is where we are seeing some distinctions in our tests.

 

[teemuk]

Take the GK 1001rb for instance vs. the mb800. What advantage would the bigger heavier rb have over the mb?

Well... what's the advantage of an apple in comparison to an orange?

Quite frankly, most often the comparisons between class-D amps and class-AB amps are made between amps that are quite different in all other aspects as well.

That being said, the main differences that we are seeing in our testing comes when we apply the burst power tests.

That applies to pretty much every amp. You'll find similar differences in burst power even when you compare class-AB amps to other class-AB amps or class-D amps to other class-D amps. There is no universal rule that class-D would be inferior in burst power delivery in comparison to class-AB, class A, G, H or whatever.

 

[tombowlus]

That applies to pretty much every amp. You'll find similar differences in burst power even when you compare class-AB amps to other class-AB amps or class-D amps to other class-D amps. There is no universal rule that class-D would be inferior in burst power delivery in comparison to class-AB, class A, G, H or whatever.

Agreed. My point is that in testing a variety of class-D, class-AB, class-G/H heads, the main differences which we are seeing between the different topologies in our measured results (regardless of brand of amp) are in the results of the burst power tests.

 

[teemuk]

That could be, but it's also a characteristic that might not introduce any significancy.

I know one guitar amp that has a burst power of about 280 watts. The continuos power is closer to only 60 watts, though. It's definitely an audible difference power-wise but we are talking about a short burst transient to which human hearing is actually very insensitive to. So any headroom most likely goes undetected and overall the only important spec we can relate to is the continuous average power.

Then we may have certain (in)famous amplifiers that boast a fairly high output power, sound unspeakably loud but actually measure far worse than the quoted specs due to internal limiting circuitry that softly distorts transients while boosting the average amplitude. These amplifiers might easily be louder and seemingly more powerful than many other amps that feature higher output power, greater burst power cababilities and greater crest factor.

So what do the specs actually tell us when it comes to instrument amps that are not usually judged by the same standards as HiFi units...?

Unfortunately, not much.

 

[seamonkey]

"FTC" or "EIA" ratings are better than nothing.

If the amp is not rated in "FTC" or another standard like "EIA", then that 100% watt number the manufacturer puts out is suspect.

And bursts over 100% are not anchored to a standard.

I certainly don't care about "peak" power in the specs of amps I look at.

 

[Warmuth]

Well... what's the advantage of an apple in comparison to an orange?

Quite frankly, most often the comparisons between class-D amps and class-AB amps are made between amps that are quite different in all other aspects as well.



That applies to pretty much every amp. You'll find similar differences in burst power even when you compare class-AB amps to other class-AB amps or class-D amps to other class-D amps. There is no universal rule that class-D would be inferior in burst power delivery in comparison to class-AB, class A, G, H or whatever.

I get what you're saying. I guess the question could be phrased in that case as "if an apple weighs 4 pounds and an orange weighs 22 and we can conceivably make an apple taste like either one, whats the point of the orange, given that smaller is preferable for a litany of reasons?". The apple, last I checked is also cheaper.

Tube amps survived, mostly in the guitar world but there are select bass amps, because some people still prefer an innate something they get from tubes. I just wondered if there would be any reason to stay with class AB or if it's just outdated tech that will be going away at some point in favor of class D entirely.

 

[ThisBass]

Did anybody ever did a voltage measurement to an amp's output when a bass instrument is plugged into the amp?

Did anybody ever scoped a sample of dynamic transients of a bass guitar signal and tried to translate the peak sample into the burst potential of a given amp?

Did anybody ever did an estimate how much portion of dynamic transients must be clipped until it's audible?

Did anybody ever gave a serious consideration why Class D amps do not need mains adpters that is twice of the rated output.

Did anybody ever considered what happens when a signal intends to be clipped:
1) like Class A or AB with a double potent power inside the net transformer
2) Class D has it not!

Did anybody ever draw a sample of a bass guitars signal from the very first initial fingers picking until the average swinging of the string?
Did anybody ever calculated the crest factor?


In my opinion the capability of a 5..10 msec.burst that provides +45% above rated power is never at all a benefit.
A conventional Class AB amp may be clipped for a time duration of 0.3 seconds until the clipping is audible. That's the deal.

 

[B-string]

I am confused as to whether you are grinding an axe or ???
BUT you may like this link http://www.talkbass.com/forum/f15/bass-frequency-waterfall-plots-what-they-mean-rigs-510749/
The FAQs/stickies are a wealth of info.

 

[ThisBass]

I am confused as to whether you are grinding an axe or ???
BUT you may like this link http://www.talkbass.com/forum/f15/bass-frequency-waterfall-plots-what-they-mean-rigs-510749/
The FAQs/stickies are a wealth of info.

I'd to claim the action is very very low ....
The initial finger peak drives subharmonic content but, that is not seen on the diagram ???

ähm,
I do not want to be grinding.

I try to to tell why conventional AB amps feel stronger then class D ever can do it with same wattage.
That's all.

 

[tombowlus]

Did anybody ever did a voltage measurement to an amp's output when a bass instrument is plugged into the amp?

Did anybody ever scoped a sample of dynamic transients of a bass guitar signal and tried to translate the peak sample into the burst potential of a given amp?

Did anybody ever did an estimate how much portion of dynamic transients must be clipped until it's audible?

Did anybody ever gave a serious consideration why Class D amps do not need mains adpters that is twice of the rated output.

Did anybody ever considered what happens when a signal intends to be clipped:
1) like Class A or AB with a double potent power inside the net transformer
2) Class D has it not!

Did anybody ever draw a sample of a bass guitars signal from the very first initial fingers picking until the average swinging of the string?
Did anybody ever calculated the crest factor?


In my opinion the capability of a 5..10 msec.burst that provides +45% above rated power is never at all a benefit.
A conventional Class AB amp may be clipped for a time duration of 0.3 seconds until the clipping is audible. That's the deal.

Yes.

 

[Jeff Scott]

Yes.

Simple, and to the point!

 

[B-string]

Yes.

Simple, and to the point!

Exactly, I just do not have the patience this week for point by point.
Most rudimentary the question of mains draw: Class A/B <60% efficiency input main draw for output power. Class D > 85% most being more efficient than 85% at power usage.

 

[tombowlus]

To clarify, I think that ThisBass raises some excellent points with his/her questions. While my response was a bit tongue in cheek, the truth is that our Technical Editor, Tom Lees, most definitely has considered many of the questions posited in ThisBass' post. Some of these questions - and the answers/insight gleaned from addressing them (or at least attempting to do so) - impacted how our bench tests were set up. In addition, some of the topics raised are slated for further discussion in some of Tom's upcoming From the Bench articles.

So, to be clear, I intended no disrespect. I think that ThisBass has raised some very valid points.

 

[agedhorse]

Did anybody ever did a voltage measurement to an amp's output when a bass instrument is plugged into the amp?

Did anybody ever scoped a sample of dynamic transients of a bass guitar signal and tried to translate the peak sample into the burst potential of a given amp?

Did anybody ever did an estimate how much portion of dynamic transients must be clipped until it's audible?

Did anybody ever gave a serious consideration why Class D amps do not need mains adpters that is twice of the rated output.

Did anybody ever considered what happens when a signal intends to be clipped:
1) like Class A or AB with a double potent power inside the net transformer
2) Class D has it not!

Did anybody ever draw a sample of a bass guitars signal from the very first initial fingers picking until the average swinging of the string?
Did anybody ever calculated the crest factor?


In my opinion the capability of a 5..10 msec.burst that provides +45% above rated power is never at all a benefit.
A conventional Class AB amp may be clipped for a time duration of 0.3 seconds until the clipping is audible. That's the deal.

Yes, some of us who design these products for a living (I have designed products with both class AB/G/H and class D amps, with both line frequency and SMPS type power supplies), and over the last 30 years of doing so have researched so many minute aspects of bass amplification that your list barely scratches the surface.

While some of your statements are partially valid, most are gross assumptions or opinions that IME are just not accurate representations of the technical facts, are partial-truths that are misleading without much more qualification.

I'll take the last one for example... who ever said that a burst rating has to be 5 or 10mSec? My own design criteria use burst periods that are relevant to the primary weighted power bandwidth. For a bass amp, that would be WAY longer than 5 or 10msec. For a high frequency amp in a biamped pro audio product, 10 or 20mSec might be entirely appropriate but you do not explain this, or WHY this is important. Additionally, your comment about a class AB amp's clipping is not audible for 0.3 sec is not IMO any more valid than a class D amp's clipping not being audible for 0.3 sec. In a properly designed amplifier, it is possible for clipping behavior to be made identical regardless of class. It is also possible for a class D amp to clip way more gracefully than a class AB amp depending on design. It all depends on the skill of the engineers in the art of non-linear signal management, I happen to have a patent in this specific area.

Your earlier arguments are just as questionable IMO.

IMO, it's important that discussions like this are brought back to a factual nature so that folks trying to learn about this technology do not get confused or misled. It's confusing stuff already, no need to muddy the waters. The more folks really understand about this, the better it is for the entire bass community. There are some other really talented designers that frequent these forums, who are at least as skilled and experienced as I am in their art of design. I openly invite their participation as this is how we all move forward in learning new technology.

 

[Duke21]

Did anybody ever did a voltage measurement to an amp's output when a bass instrument is plugged into the amp?

Did anybody ever scoped a sample of dynamic transients of a bass guitar signal and tried to translate the peak sample into the burst potential of a given amp?

Did anybody ever did an estimate how much portion of dynamic transients must be clipped until it's audible?

Did anybody ever gave a serious consideration why Class D amps do not need mains adpters that is twice of the rated output.

Did anybody ever considered what happens when a signal intends to be clipped:
1) like Class A or AB with a double potent power inside the net transformer
2) Class D has it not!

Did anybody ever draw a sample of a bass guitars signal from the very first initial fingers picking until the average swinging of the string?
Did anybody ever calculated the crest factor?

In my opinion the capability of a 5..10 msec.burst that provides +45% above rated power is never at all a benefit.
A conventional Class AB amp may be clipped for a time duration of 0.3 seconds until the clipping is audible. That's the deal.

I'm quite sure TC electronics looked on this as they designed their amps according to your last assumption.

 

[ThisBass]

Yes, some of us who design these products for a living (I have designed products with both class AB/G/H and class D amps, with both line frequency and SMPS type power supplies), and over the last 30 years of doing so have researched so many minute aspects of bass amplification that your list barely scratches the surface.

While some of your statements are partially valid, most are gross assumptions or opinions that IME are just not accurate representations of the technical facts, are partial-truths that are misleading without much more qualification.

I'll take the last one for example... who ever said that a burst rating has to be 5 or 10mSec? My own design criteria use burst periods that are relevant to the primary weighted power bandwidth. For a bass amp, that would be WAY longer than 5 or 10msec. For a high frequency amp in a biamped pro audio product, 10 or 20mSec might be entirely appropriate but you do not explain this, or WHY this is important. Additionally, your comment about a class AB amp's clipping is not audible for 0.3 sec is not IMO any more valid than a class D amp's clipping not being audible for 0.3 sec. In a properly designed amplifier, it is possible for clipping behavior to be made identical regardless of class. It is also possible for a class D amp to clip way more gracefully than a class AB amp depending on design. It all depends on the skill of the engineers in the art of non-linear signal management, I happen to have a patent in this specific area.

Your earlier arguments are just as questionable IMO.

IMO, it's important that discussions like this are brought back to a factual nature so that folks trying to learn about this technology do not get confused or misled. It's confusing stuff already, no need to muddy the waters. The more folks really understand about this, the better it is for the entire bass community. There are some other really talented designers that frequent these forums, who are at least as skilled and experienced as I am in their art of design. I openly invite their participation as this is how we all move forward in learning new technology.

+1

In general I estimate the Peak To RMS (or crest) of any Bass Signal to approximate +10dB.
I know this value can be even dramatically more, or a little bit less. Depending on the instrument the strings and the players technique.

What do you think? How strong do you estimate (or measured) the crest?

As an example I said it is something approximate +10dB.
Amp power 500 Watt (or a little bit more with burst capability).
If I do not want the amp to be clipped I need a little bit headroom, let's say +3dB which is 150 Watt.
Depending on the players ability to play most of the time steady +3dB is not really very much headroom.

The RMS for the signal calculates then to little 15 Watt rms.
That's far to less to be loud enough in many situations.

Maybe my estimition +10 dB Peak To RMS is wrong but I don't think it is so.
what do you think?

On the other side it is not a problem to let the signal peaks clip a little bit cause it's not audible anyway.

My estimation is circa 0.3 second duration till it's audible for uncompressed signals.
Depending on sound and frequency response of the cab this value may vary a little bit I know.

What do you think? Or what did you measured?

Translated into an amp power 300 Watt this would be approximate 100 Watt rms. 100 watt average that includes 200 watt clipping peak content.

What do you think or what did you measured?

There is some area below where the clipping behavior can be noticed as a smooth sound compressing. I think this would be the most common situation when many players think there cabs get enough power to sound really good.
I know there are many players all around that feel happy with this. Or an the other side, many players argue their cabs sound pretty good if the power is big enough that goes to the cabs.
Some players argue there would be an improved and excellent dynamic when the amp is turned to higher output cause of the smooth feeling of playing notes.

I'd say it different. Many players feel happy if the amp produces enough non audible clipping content and gives therefore a smooth compress to the signal.


What do you think? What's your experience?


I hope you give some more detailed information to the burst capability. Or give more detailed information how the clipping scenario in an class D works.

I know it is not necessary for any amp to provide full power for the duration of seconds. But it has to be long enough for the transients.

There are some professional amps on the market that provide nearly twice the power at the output than it is received on the socket.
It's a clear thing that this is never possible with a sine wave but, it's possible cause of the nature of audio signals.

What's your opinion?
Is it telling the truth to rate a professional amp to 5.000 watt and spec it to only 1.000 watt power consumption at the socket?

How do you estimate and rate the power consumption at the socket for your amps?


I never had the ability to hear how it really sounds whenever a PWM signal comes up to it's ending and therefore starts to clip.
All I know is it should be terrible noise.

Do you have any examples?

Some limiters on the market delay the signal a little bit (non audible) to provide enough time for reduction.
The benefit is a nearly perfect limited signal.

Is this limiter concept a essential part of your class D amps?

If it is not then how do you manage the scenario of very powerful dynamics in the signal?


A lot of questions I know.
But I think it's valuable to know a little bit more how the PWM concepts really do their work.

Last question:
how do you estimate/calculate the switchmode power supply capacity for a class D bass amplifier, let's say for a 500 Watt rated amp?
Do you design your own class D power modules or do you buy them like some other manufactures do it?
What do you think about B&O or Powersoft or Hoellstern?

I think the B&O Ice power module's heating capacity is designed for only 1/8 of rated output.
A little bit too weak for an application like a bass guitar amplifier I'd suppose.
What's your opinion to that design?

 

[agedhorse]

+1

In general I estimate the Peak To RMS (or crest) of any Bass Signal to approximate +10dB.
I know this value can be even dramatically more, or a little bit less. Depending on the instrument the strings and the players technique.What do you think? How strong do you estimate (or measured) the crest?

Can vary greatly, but first you have to compare apples to apples. When comparing peak to RMS values for identical signals (sine wave) the peak value is already 3dB higher. You need to compare the peak to the peak. In my experience this will range from 6dB to ~9dB with slapping being a special case for a variety of reasons. Now, when calculating this ratio, you need to remember that any amp that is rated for say 600 watts RMS will also deliver 1200 watts peak (it's purely a math conversion)

As an example I said it is something approximate +10dB. Amp power 500 Watt (or a little bit more with burst capability). If I do not want the amp to be clipped I need a little bit headroom, let's say +3dB which is 150 Watt. Depending on the players ability to play most of the time steady +3dB is not really very much headroom.

Let me correct the math here. First, we will look at peak values (for consistant math), so your 500 watt RMS amp is actually 1000 watts peak. Let's use 9dB for headroom (easy for math) the "average" of the PEAK power of the signal with 9dB crest factor would be 125 watts, or ~75 watts RMS. Note that I am being a bit liberal with the termonology because we are deviating from the sine wave definitions.

The RMS for the signal calculates then to little 15 Watt rms. That's far to less to be loud enough in many situations. Maybe my estimition +10 dB Peak To RMS is wrong but I don't think it is so. what do you think?

Once the math is adjusted, I think we are in a more realistic area now. 10dB is an awful lot of dynamic range for bass and highlights why often compressors or compression like algorithems that are used in bass amp design in order to maoderate the peak to average ratio to more pleasing or more useful (in a band setting) level. This is also part of a player's technique, how to fit into a mix, find the pocket and especially stay there.

On the other side it is not a problem to let the signal peaks clip a little bit cause it's not audible anyway. My estimation is circa 0.3 second duration till it's audible for uncompressed signals. Depending on sound and frequency response of the cab this value may vary a little bit I know. What do you think? Or what did you measured?

There are 2 things going on here the need to be split apart. One is the basic sound or texture of clipping which can be manipulated through design to be part of the overall personality of the amp. I think the SVT is probably the best example of this. It's one of the things that has made this amp such a desireable piece for so many years (if the weight isn't an issue of course). The other part of this is the behavior of the clipping. It can clip in a graceful, pleasing way or it can clip in a bad way as well. Clipping, when not well managed through the design process CAN (but doesn't have to) sound terrible. It's the artifacts such as chattering, phase shifted current clipping, latching and sticking, etc. are really defects in design rather than directly the result of clipping.

Translated into an amp power 300 Watt this would be approximate 100 Watt rms. 100 watt average that includes 200 watt clipping peak content. hat do you think or what did you measured?

This can be all over the map IMO.

There is some area below where the clipping behavior can be noticed as a smooth sound compressing. I think this would be the most common situation when many players think there cabs get enough power to sound really good.
I know there are many players all around that feel happy with this. Or an the other side, many players argue their cabs sound pretty good if the power is big enough that goes to the cabs.
Some players argue there would be an improved and excellent dynamic when the amp is turned to higher output cause of the smooth feeling of playing notes.

I'd say it different. Many players feel happy if the amp produces enough non audible clipping content and gives therefore a smooth compress to the signal.


What do you think? What's your experience?

IMO, this is one of the major parts of any amp design that makes one amp better for one type of player than another. When the clipping is combined with limiting, compression, multi-band compression and limiting, frequency content ratio adjustments, etc. this becomes a tool to establish a variety of personalities.

I hope you give some more detailed information to the burst capability. Or give more detailed information how the clipping scenario in an class D works.

I know it is not necessary for any amp to provide full power for the duration of seconds. But it has to be long enough for the transients.

Burst ratings are done in a variety of ways (some for good reasons, some for marketing reasons) but essentially look at the ability of an amp to handle peak power above and beyond what it can do in steady state operation. Since bass signals have dynamics, an amp that can deliver 500 watts steady state RMS will also deliver 1000 watts peak (in steady state operation) but may deliver 1500 or 1700 watts peak for a defined period of time in non-steady state operation. This is part of the feel under playing conditions that can make on amp feel bigger, or more open or dynamic than another with the same continuous rating. The question of the time period and repitition rate of the burst test signal depends on the designer's goal for the feel of the amp. Generally for bass, I like to look at something around 50-100 msec, enough for several periods of a 50Hz signal to be reproduced before the decay envelope of the signal falls. This is where understanding the nature of the bas signal is helpful in extracting feel from an amp design. 5-10 mSec is pretty much useless (IMO) because it doesn't allow for even 1/2 of a period at 50Hz (unless this is part of the specific feel the designer is after).

There are some professional amps on the market that provide nearly twice the power at the output than it is received on the socket.
It's a clear thing that this is never possible with a sine wave but, it's possible cause of the nature of audio signals.

What's your opinion?

The power consumption stated on the amp is governed by several ruling bodies (EU/CE, UL, CSA, etc.) but in general mus be no less that the power consumption at 1/8-power. It is to provide a general average power consumption for electrical calculations. It includes whatever inefficiencies the amp reflects. I find that for bass, 1/4-power is more representative for many playing styles, however.

Is it telling the truth to rate a professional amp to 5.000 watt and spec it to only 1.000 watt power consumption at the socket?

Yes, it is, a pro audio amp is generally used on a full range signal where 1/8-power is pretty reasonable. Sub amps will generally be closer to 1/4-power. This is why most pro audio amps include power draw versus duty cycle charts in their documentation.

How do you estimate and rate the power consumption at the socket for your amps?

It depends on the specific product. Generally speaking we use ~1/4 power on bass amps and ~1/8-power on acoustic guitar amps, reflecting the typical dynamics variations and frequency density content.

I never had the ability to hear how it really sounds whenever a PWM signal comes up to it's ending and therefore starts to clip. All I know is it should be terrible noise.

Do you have any examples?

If the designer alows the PWM process to be interrupted or corrupted then it certainly can sound terrible. Most designers of quality class D products go to great pains to isolate or modify the PWM process from the effects of clipping (or specifically running out of dynamic range). We do this in every one of our amps, in different ways depending on the feel we are trying to achieve.

Some limiters on the market delay the signal a little bit (non audible) to provide enough time for reduction. The benefit is a nearly perfect limited signal.

Look ahead type limiters indeed offer this advantage, and is one of the benefits of digital processing to allow time for more complex algorithems. Unfortunately it requires an A/D conversion, DSP and then a D/A conversion whcih adds cost, complexity, a non-analog signal path and possibly additional unwanted latency.

Is this limiter concept a essential part of your class D amps?

Mo, everything is done in the analog domain.

If it is not then how do you manage the scenario of very powerful dynamics in the signal?

I can't get into the specifics, I can say that it's all well addressed within the complete design... ie. all parts are designed to work together to support each other and not generate negative artifacts.

I am splitting your post here, see the next entry.

 

[agedhorse]

A lot of questions I know.
But I think it's valuable to know a little bit more how the PWM concepts really do their work.

The "basic" operation is as follows:

The analog signal is compared to a clocked triangle reference wave and this generates a PWM signal. The PWM signal varies it's width and repitition according to the analog signal. This signal is then level shifted from a logic level to a power amp rail level say from 5 volts PWM (peak) to 100 volts analog audio peak. The signal is then low passed in a high powered low pass filter which extracts the analog audio from the PWM signal. Of course this is a gross oversimplification, but it's the basics.

Last question:
how do you estimate/calculate the switchmode power supply capacity for a class D bass amplifier, let's say for a 500 Watt rated amp?

It depends on what is desired with regards to dynamics reproduction, with what kind of burst performance that is desired, what the overall efficiency of the amp is and how the amp reacts to variations in the supply (ie. power supply rejection). There are both average and peak ratings that are considered. I can't go into specifics here.

Do you design your own class D power modules or do you buy them like some other manufactures do it?
What do you think about B&O or Powersoft or Hoellstern?

I have designed some class D amps but they are difficult to do well, require enormous resources and the cost of testing and agency approvals can be staggeringly high. Anybody who thinks that designing a high quality, reliable class D amp has obviously never done so. I have worked with B&O since they introduced their products. I also work in the pro audio industry and have enormous respect for what they have accomplished. They have a group of highly skilled and specialized engineers working on these designs. I am familiar with Powersoft and others but not to the same depth so i can't comment. I am not familiar with Hoellstern.

I think the B&O Ice power module's heating capacity is designed for only 1/8 of rated output.
A little bit too weak for an application like a bass guitar amplifier I'd suppose.
What's your opinion to that design?

I can't go into specifics, but the data sheet specs are misleading IF you understand what's going on between the lines of their designs. I have worked closely with their engineers when developing our products and because I have a comprehensive understanding of their designs I know how to exploit specific performance enhancements that would not be obvious to folks who have not designed products of this type. IMO, the designs are very thorough, and brilliantly conceived provided you know what you are looking at and understand why they are done the way they are. It's not at all obvious.

 

[Duke21]

The "basic" operation is as follows:

The analog signal is compared to a clocked triangle reference wave and this generates a PWM signal. The PWM signal varies it's width and repitition according to the analog signal. This signal is then level shifted from a logic level to a power amp rail level say from 5 volts PWM (peak) to 100 volts analog audio peak. The signal is then low passed in a high powered low pass filter which extracts the analog audio from the PWM signal. Of course this is a gross oversimplification, but it's the basics.

It depends on what is desired with regards to dynamics reproduction, with what kind of burst performance that is desired, what the overall efficiency of the amp is and how the amp reacts to variations in the supply (ie. power supply rejection). There are both average and peak ratings that are considered. I can't go into specifics here.

I have designed some class D amps but they are difficult to do well, require enormous resources and the cost of testing and agency approvals can be staggeringly high. Anybody who thinks that designing a high quality, reliable class D amp has obviously never done so. I have worked with B&O since they introduced their products. I also work in the pro audio industry and have enormous respect for what they have accomplished. They have a group of highly skilled and specialized engineers working on these designs. I am familiar with Powersoft and others but not to the same depth so i can't comment. I am not familiar with Hoellstern.

I can't go into specifics, but the data sheet specs are misleading IF you understand what's going on between the lines of their designs. I have worked closely with their engineers when developing our products and because I have a comprehensive understanding of their designs I know how to exploit specific performance enhancements that would not be obvious to folks who have not designed products of this type. IMO, the designs are very thorough, and brilliantly conceived provided you know what you are looking at and understand why they are done the way they are. It's not at all obvious.

I'm impressed by your answering here! Really deep stuff.