• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Marchaudio P422 Stereo Power Amplifier Renewed Review

sarumbear

Master Contributor
Forum Donor
Joined
Aug 15, 2020
Messages
7,604
Likes
7,313
Location
UK
It is an empirical value. If you pick much lower number, due to variability of noise+distortion, you get different run to run numbers. At 1% there is enough distortion that the computed power at that distortion has very small variation. I spent good bit of time testing before arriving at this value (and used 0.1% at one point).
I understand. Thank you for explaining.
 

PeteL

Major Contributor
Joined
Jun 1, 2020
Messages
3,303
Likes
3,838
That’s the reason d’être of ASR, isn’t it, to out those naughty manufacturers.


That’s when you use the asterisks.


I was simply hoping that as ASR making SINAD a standard it could create a modern standard that is fit for the era and current market. Other than at powers they are not designed for (but sometimes advertised as by marketers) all modern amplifiers operates at less than 1% distortion.

To repeat, I’m simply writing my views, not critiquing anyone.
Let’s just say that if an amp is specced say 200 W 4 ohm, 1% THD. if you never demand more than 200 W, You’ll never hear any disortion, so in my book it may well be a conservative value and not a high one. You will hit that 1% on short peaks. It’s already disputed that 1% distortion is audible at all, as an operating spec, for some peaks here and there it’s totally impossible to hear. If the loudest demand in your listening session generate 1% distortion, it will be totally transparent, and I think you can say that you are not really pushing the amp out of it’s linear operation.
 

Holmz

Major Contributor
Joined
Oct 3, 2021
Messages
2,018
Likes
1,241
Location
Australia

I was simply hoping that as ASR making SINAD a standard it could create a modern standard that is fit for the era and current market. Other than at powers they are not designed for (but sometimes advertised as by marketers) all modern amplifiers operates at less than 1% distortion.

Maybe they operate at <1% for the majority of the music, but the crest factor sort of dictates what the RMS value is versus the peaks.
So if 99% os the music is under that 200W, then the %distortion would be maybe 0.01%
But the 1% of the peaks being clipped are severely distorted.
It is still 1%, but the “clipping 1%” would be a lot different than 1% from say harmonic distortion covering both the loud and quiet parts…

Maybe a SINAD spec at 90% of rated power makes sense? But we already often see distortion values at 1W RMS, which is where a lot of systems are playing.
And that routinely gives 20w to 40w peaks. With crest factors of 13-16dB.
 

Billy Budapest

Major Contributor
Forum Donor
Joined
Oct 11, 2019
Messages
1,810
Likes
2,674
It is an empirical value. If you pick much lower number, due to variability of noise+distortion, you get different run to run numbers. At 1% there is enough distortion that the computed power at that distortion has very small variation. I spent good bit of time testing before arriving at this value (and used 0.1% at one point).
Plus, measuring “max” power at 1% distortion (or what some people say, defining clipping at 1% distortion) is a long-standing audio convention. I don’t know who did it first—Julian Hirsch, perhaps? In any event, it’s a convenient fiction. Conceivably, you could measure power at a higher or lower arbitrary number as Amir stated.
 

Hooch500

Member
Joined
Jun 17, 2022
Messages
7
Likes
3
After massive failure with my previous review(https://audiosciencereview.com/foru...dio-p422-stereo-power-amplifier-review.35969/), I got great technical supports from ASR, Speaker gallery of DC inside, and Marchaudio.
(Detail for the appearance, see previous review post.)

Disclaimer
This is a review and undetailed measurements of the Marchaudio P422 stereo class D power amplifier. It is on kind loan from myself and costs US $1,988.68(P422 $1,675.21 + shipping to South Korea $132.68 + import tax $180.79) as previous

But this time, Marchaudio sent one low power reference load for free.
As 3rd party reviewer, I absolutely guarantee that there was no ask of change direction of review, and the supports has no influence on the review.

P422 is Purifi 1ET400A based amp with custom designed Power supply and buffer board.

Conditions
View attachment 229886
Eventually, with AUX-0040 Filter!

and B series APx525 analyzer with AG52 as previous.

When AP connected to Class-D AMP without filter, AP can not fix suitably level, you can hear endless clicking sounds of relays.

Class-D produce signal at switching frequency(far above target measurement frequency range) so AP can't measure them. <- what I've experienced and got bunch of advices from ASR, Marchaudio

that's why I add low noise high linearity low pass filter like AUX-0040

BUT, If you are normal user, YOU NEED TO KNOW that frequency of switching is so high so that signal has no acoustic influence. Of course don't need pricey filters too.
You can simply take them as just AP's issue at all.

Also, Signal chain for the measurement has improved.
APx525 Output - Measurement Target Amp - AUX-0040 Filter - APx525 Input

And the loads
View attachment 229887View attachment 229888View attachment 229889
last time, we saw magnificent non-linearity from low heat capacity load.

This time, I got 8 x 1000 W 16 Ohms

to build
2 Ch 4000 W 4 Ohms
1 Ch 8000 W 2 Ohms
loads.

To validate new dummies

View attachment 229890View attachment 229891
I compared them with Reference load from March audio.

You can check the load in the lower right corner of the photo.

Let's see.

View attachment 229895
APx525 Output - AUX-0040 Filter - APx525 Input
Self check.

Filter works fine.

It shows the limitation of the analyzer.
Near to 109 dB we can't see certainty, Just can encounter huge errors.

so above 100 dB, I can say "We are looking at the measuring equipment itself, not the characteristics of the subject."
that case Actual values can be as good as a few dB or more.

BTW, I have to say sorry about FFT
I should display FFT window with dBrA(peak at 0 dB) not dBV(1 kHz peak is not at 0 dB).
Sorry for the inconvenience.

View attachment 229896
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven
View attachment 229897
5 W 4 Ohm
P422 amp, March's reference load Ch 1, New Load Ch 2 4 Ohm both Ch driven

We can read this result as
1.March's reference load works slightly better than new loads but still reliable.
2.P422 Amp's performance is above APx525 analyzer(We are almost seeing AP itself).
3.P422 Amp has SINAD least over 105dB, I can make vague estimate about 106 to 107 dB(Need for Apx555/Other high-resolution DAC and ADC).

But this high SINAD, you can't hear any THD+N with that hot 1kHz signal, even at the 0dB Ideal full-band 0dB anechoic chamber.
Nowadays, many of well designed electronics are way beyond human hearing.
View attachment 229898
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

the Gain
customized spec.
welcome to flat world.
View attachment 229899
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

Frequency response
Can you see?

View attachment 229900

Look at the scale
Even with 0.1 dB almost no deviation flaaaaaaaaaaaaaaaaaaat.
View attachment 229901
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

Cross talk
Meaningless!
View attachment 229902
5 W 4 Ohm
P422 amp, New Load, 4 Ohm

Cross-Cross talk
changed Ch 1 and Ch 2
Meaningless! x 2
View attachment 229903

5 W 4 Ohm
P422 amp, New Load, 4 Ohm

Many of Crosstalk
changed Ch 1 and Ch 2 and so son
Meaningless! x 3

It's not P422 AMP's Crosstalk
It's load and analyzers Crosstalk

Again, P422 Amp's performance is above the analyzer+load chaine.


We could only check P422 has normal crosstalk, It indicates that it was made normally(without problem = fine).

IMHO, P422 has very well suppressed crosstalk.

View attachment 229904
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

Multitone with dBrA

WOW, look great!

wait, did you see?
My mistake, Averages was only 1 instead of 8.
What a shame.

Floor could be much lower. I'll add multitone Averages : 8 next update.

View attachment 229905
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

Same Multitone with dBV
It's cumbersome to look at because the position of peak is not 0dB.

I did same thing all above this. SORRY!

Finally

Main Dish for Power amp measurements

View attachment 229906View attachment 229907View attachment 229908


This is the spec sheet for the Purifi 1ET400A EIGENTAKT amplifier unit.
Since the P422 is made based on this unit, the better the power supply and buffer board are made, the closer to the above performance results,
The worse the performance, the worse the results will be.

to Read this graph(level sweep)
-The higher the worse.(THD+N ratio)
-The farther to the right is the higher the output.(Output Power)

When the output is very low, the signal itself is so small that it is affected by external factors, so it shows a rather high noise + distortion level. As the output gradually increases, the signal strength becomes stronger, the purity of the signal increases, and when the finally reached to output limits, the noise and distortion increase rapid and sharp.

A very tight standard takes the performance as the effective output until the point where the sudden increase in noise & distortion begins (the inflection point)
As a general industry standard 1% (little loose), and for audio, 0.1% is used as a standard.


THD+N Ratio value of lowest point depends on performance limitation of analyzer, so pay attention to the location of the inflection point and the power(level).

with impedance
-The lower, the harder to drive.
Amp need to pump more current to load.

-The higher, the easier to drive.
Amp do not need to pump much current to load.

Then let's see P422 @ 4 Ohm

View attachment 229909

Look nice.
Let's compare.

View attachment 229909
View attachment 229911

Way higher output than Benchmark AHB 2

View attachment 229909
View attachment 229913

Audiophonics HPA-S400ET shows almost identical value.
Both amp did a nice job.

View attachment 229909
View attachment 229915


Lm Audio MIN400A shows a little quick rise than P422, HPA-S400ET.
It appears to be suffering from Power supply units limitation.

I know, all of us don't wanna see boring 4 Ohms.
2 Ohms for real!

View attachment 229916

P422 amp, New Load, 2 Ohm single Ch driven
???!

What?
View attachment 229917


Almost 460 W of Clean Pure Power at 2 Ohms!
460 W is just an inflection point!!!!!

@ 0.1% 515 W - 527 W ??!
@ 1% 554 W - 576 W ????!?!

AWESOME!
Power supply did great job!
We can't get closer to amp units limits before Reach power supply's limits with both Ch driven(1000 W rated, but in fact the PSU can handle 1100 W well).

Even with KEF R3, this amp can handle them with ease.

View attachment 229920
I'm happy to recommend the Marchaudio P422 Stereo Amplifier.

Although the performance of my measuring equipment was insufficient, so I could not completely measure the amplifier that exceeded the equipment performance, but I could see that the performance of the P422 amplifier was equal to or exceeded that of the APx525 analyzer.

Again, there were some obvious mistakes, but since it is not too difficult to judge the overall performance, I think this measurement is very valid.

I will continue to improve our measuring procedures and equipment, and really honored to post my first valid review. It would not have been possible without the support of the 'DC inside speaker gallery' members who fully supported me and the strong technical supports from ASR members, and I would like to express my special thanks to March Audio for providing the reference dummy load.

(before measurements)Meaningless subjective assessment with my own iron ears(Just cry).

It sound like crystal clear.
and I close my eyes, soon It disappears into the air feels like amp doesn't exist.

Even Otaku-music(hard to drive), it drives well.

(After measurements)
If you have P422, you can hear crystal clear Otaku music with endless cry
Vanishingly low distortion. Perhaps better suited to Chiroptologists given no human ear is ever going to hear the difference, not least because no speakers exist that can resolve the differences between
After massive failure with my previous review(https://audiosciencereview.com/foru...dio-p422-stereo-power-amplifier-review.35969/), I got great technical supports from ASR, Speaker gallery of DC inside, and Marchaudio.
(Detail for the appearance, see previous review post.)

Disclaimer
This is a review and undetailed measurements of the Marchaudio P422 stereo class D power amplifier. It is on kind loan from myself and costs US $1,988.68(P422 $1,675.21 + shipping to South Korea $132.68 + import tax $180.79) as previous

But this time, Marchaudio sent one low power reference load for free.
As 3rd party reviewer, I absolutely guarantee that there was no ask of change direction of review, and the supports has no influence on the review.

P422 is Purifi 1ET400A based amp with custom designed Power supply and buffer board.

Conditions
View attachment 229886
Eventually, with AUX-0040 Filter!

and B series APx525 analyzer with AG52 as previous.

When AP connected to Class-D AMP without filter, AP can not fix suitably level, you can hear endless clicking sounds of relays.

Class-D produce signal at switching frequency(far above target measurement frequency range) so AP can't measure them. <- what I've experienced and got bunch of advices from ASR, Marchaudio

that's why I add low noise high linearity low pass filter like AUX-0040

BUT, If you are normal user, YOU NEED TO KNOW that frequency of switching is so high so that signal has no acoustic influence. Of course don't need pricey filters too.
You can simply take them as just AP's issue at all.

Also, Signal chain for the measurement has improved.
APx525 Output - Measurement Target Amp - AUX-0040 Filter - APx525 Input

And the loads
View attachment 229887View attachment 229888View attachment 229889
last time, we saw magnificent non-linearity from low heat capacity load.

This time, I got 8 x 1000 W 16 Ohms

to build
2 Ch 4000 W 4 Ohms
1 Ch 8000 W 2 Ohms
loads.

To validate new dummies

View attachment 229890View attachment 229891
I compared them with Reference load from March audio.

You can check the load in the lower right corner of the photo.

Let's see.

View attachment 229895
APx525 Output - AUX-0040 Filter - APx525 Input
Self check.

Filter works fine.

It shows the limitation of the analyzer.
Near to 109 dB we can't see certainty, Just can encounter huge errors.

so above 100 dB, I can say "We are looking at the measuring equipment itself, not the characteristics of the subject."
that case Actual values can be as good as a few dB or more.

BTW, I have to say sorry about FFT
I should display FFT window with dBrA(peak at 0 dB) not dBV(1 kHz peak is not at 0 dB).
Sorry for the inconvenience.

View attachment 229896
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven
View attachment 229897
5 W 4 Ohm
P422 amp, March's reference load Ch 1, New Load Ch 2 4 Ohm both Ch driven

We can read this result as
1.March's reference load works slightly better than new loads but still reliable.
2.P422 Amp's performance is above APx525 analyzer(We are almost seeing AP itself).
3.P422 Amp has SINAD least over 105dB, I can make vague estimate about 106 to 107 dB(Need for Apx555/Other high-resolution DAC and ADC).

But this high SINAD, you can't hear any THD+N with that hot 1kHz signal, even at the 0dB Ideal full-band 0dB anechoic chamber.
Nowadays, many of well designed electronics are way beyond human hearing.
View attachment 229898
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

the Gain
customized spec.
welcome to flat world.
View attachment 229899
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

Frequency response
Can you see?

View attachment 229900

Look at the scale
Even with 0.1 dB almost no deviation flaaaaaaaaaaaaaaaaaaat.
View attachment 229901
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

Cross talk
Meaningless!
View attachment 229902
5 W 4 Ohm
P422 amp, New Load, 4 Ohm

Cross-Cross talk
changed Ch 1 and Ch 2
Meaningless! x 2
View attachment 229903

5 W 4 Ohm
P422 amp, New Load, 4 Ohm

Many of Crosstalk
changed Ch 1 and Ch 2 and so son
Meaningless! x 3

It's not P422 AMP's Crosstalk
It's load and analyzers Crosstalk

Again, P422 Amp's performance is above the analyzer+load chaine.


We could only check P422 has normal crosstalk, It indicates that it was made normally(without problem = fine).

IMHO, P422 has very well suppressed crosstalk.

View attachment 229904
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

Multitone with dBrA

WOW, look great!

wait, did you see?
My mistake, Averages was only 1 instead of 8.
What a shame.

Floor could be much lower. I'll add multitone Averages : 8 next update.

View attachment 229905
5 W 4 Ohm
P422 amp, New Load, 4 Ohm both Ch driven

Same Multitone with dBV
It's cumbersome to look at because the position of peak is not 0dB.

I did same thing all above this. SORRY!

Finally

Main Dish for Power amp measurements

View attachment 229906View attachment 229907View attachment 229908


This is the spec sheet for the Purifi 1ET400A EIGENTAKT amplifier unit.
Since the P422 is made based on this unit, the better the power supply and buffer board are made, the closer to the above performance results,
The worse the performance, the worse the results will be.

to Read this graph(level sweep)
-The higher the worse.(THD+N ratio)
-The farther to the right is the higher the output.(Output Power)

When the output is very low, the signal itself is so small that it is affected by external factors, so it shows a rather high noise + distortion level. As the output gradually increases, the signal strength becomes stronger, the purity of the signal increases, and when the finally reached to output limits, the noise and distortion increase rapid and sharp.

A very tight standard takes the performance as the effective output until the point where the sudden increase in noise & distortion begins (the inflection point)
As a general industry standard 1% (little loose), and for audio, 0.1% is used as a standard.


THD+N Ratio value of lowest point depends on performance limitation of analyzer, so pay attention to the location of the inflection point and the power(level).

with impedance
-The lower, the harder to drive.
Amp need to pump more current to load.

-The higher, the easier to drive.
Amp do not need to pump much current to load.

Then let's see P422 @ 4 Ohm

View attachment 229909

Look nice.
Let's compare.

View attachment 229909
View attachment 229911

Way higher output than Benchmark AHB 2

View attachment 229909
View attachment 229913

Audiophonics HPA-S400ET shows almost identical value.
Both amp did a nice job.

View attachment 229909
View attachment 229915


Lm Audio MIN400A shows a little quick rise than P422, HPA-S400ET.
It appears to be suffering from Power supply units limitation.

I know, all of us don't wanna see boring 4 Ohms.
2 Ohms for real!

View attachment 229916

P422 amp, New Load, 2 Ohm single Ch driven
???!

What?
View attachment 229917


Almost 460 W of Clean Pure Power at 2 Ohms!
460 W is just an inflection point!!!!!

@ 0.1% 515 W - 527 W ??!
@ 1% 554 W - 576 W ????!?!

AWESOME!
Power supply did great job!
We can't get closer to amp units limits before Reach power supply's limits with both Ch driven(1000 W rated, but in fact the PSU can handle 1100 W well).

Even with KEF R3, this amp can handle them with ease.

View attachment 229920
I'm happy to recommend the Marchaudio P422 Stereo Amplifier.

Although the performance of my measuring equipment was insufficient, so I could not completely measure the amplifier that exceeded the equipment performance, but I could see that the performance of the P422 amplifier was equal to or exceeded that of the APx525 analyzer.

Again, there were some obvious mistakes, but since it is not too difficult to judge the overall performance, I think this measurement is very valid.

I will continue to improve our measuring procedures and equipment, and really honored to post my first valid review. It would not have been possible without the support of the 'DC inside speaker gallery' members who fully supported me and the strong technical supports from ASR members, and I would like to express my special thanks to March Audio for providing the reference dummy load.

(before measurements)Meaningless subjectiveassessment with my own iron ears(Just cry).

It sound like crystal clear.
and I close my eyes, soon It disappears into the air feels like amp doesn't exist.

Even Otaku-music(hard to drive), it drives well.

(After measurements)
If you have P422, you can hear crystal clear Otaku music with endless cry
 
OP
thin bLue

thin bLue

Senior Member
Reviewer
Joined
Jul 29, 2019
Messages
352
Likes
1,155
Vanishingly low distortion. Perhaps better suited to Chiroptologists given no human ear is ever going to hear the difference, not least because no speakers exist that can resolve the differences between
Thanks for cheering me!

But, as an armchair topologist, I'm sure we can see fairly large difference between level sweep datas.

For SINAD, yes we can't hear any of them.
But most of us wanna line them(AMPs, DACs and etc.) up.
 
OP
thin bLue

thin bLue

Senior Member
Reviewer
Joined
Jul 29, 2019
Messages
352
Likes
1,155
To the OP: what was the amplifier gain on the P422 ?
20.5 db via XLr or 26.5 via RCA ?
20220628_185923.jpg

This is custom gain of mine(If you want to change the gain settings, you can ask when you order. Mail to March audio for the details!)

Measurements were at 14.5 dB mode(without jumper)
 
OP
thin bLue

thin bLue

Senior Member
Reviewer
Joined
Jul 29, 2019
Messages
352
Likes
1,155
How does this compare to the Purifi reference design?
According to Marchaudio, Bruno Putzeys stated; “your input buffer is more transparent than our Eval1 board”
(I heard "Bruno Putzeys said that at last HIGH END Munich show, May 2022")

As I know Bruno Putzeys is co-founder of Purifi and amp architect.

So, We could expect P422 has at least equal or better performance than Eval bord.
but, March has input RF filter so can reduce possible IMD problems with Class-D that's the strong point of P422
So, could be feel happier xD.
 
Last edited:
OP
thin bLue

thin bLue

Senior Member
Reviewer
Joined
Jul 29, 2019
Messages
352
Likes
1,155
Very nice and thorough measurement report and review.
With all respect, and it might be a native langage thing, but this is a bit of a odd way to explain this phenomenon, or at the very least unclear.

The noise + distortion "level" is not higher at low volume. It's a ratio it's just closer relative to the signal, not in absolute terms. I would also not say that the "purity" of the signal increase or at least specify that by signal you mean the total signal + noise summation. But the purity of the signal don't increase, it just become farther from noise, in fact it may contain a bit more distortion as you bring it up, or no more, but not less. All this really say is that the sum of noise and distortion don't increase as fast as the signal level.

Nice feedback!!!!
Your description is way better than mine!

Thank you for giving me chance to improve my review!

I think, I should reconsider my wrong and vague explain. (in fact, I wrote and uploaded this review about AM 5:30 so my poor English got even worse )
 

Rick Sykora

Major Contributor
Forum Donor
Joined
Jan 14, 2020
Messages
3,513
Likes
7,017
Location
Stow, Ohio USA
According to Marchaudio, Bruno Putzeys stated; “your input buffer is more transparent than our Eval1 board”
(I heard "Bruno Putzeys said that at last HIGH END Munich show, May 2022")

As I know Bruno Putzeys is co-founder of Purifi and amp architect.

So, We could expect P422 has at least equal or better performance than Eval bord.
but, March has input RF filter so can reduce possible IMD problems with Class-D that's the strong point of P422
So, could be feel happier xD.

Was Bruno’s “more transparent” purely based on measurements?

Just have to wonder how to value that better transparency as seems unlikely it is audibly so.
 
OP
thin bLue

thin bLue

Senior Member
Reviewer
Joined
Jul 29, 2019
Messages
352
Likes
1,155
Was Bruno’s “more transparent” purely based on measurements?

Just have to wonder how to value that better transparency as seems unlikely it is audibly so.

IMHO, it might be a few dB of SINAD differences between measurements. But I don't have information to judge so I can't make any answer about that.

But we can ask them!
 

Ajax

Active Member
Forum Donor
Joined
Aug 14, 2018
Messages
248
Likes
802
Location
Byron Bay, Australia
After being an avid amateur for 40 years, I can assure you that it has never been more important to have the knowledge to understand audio measurements, especially with amplifiers of this calibre at just US $1700 and it's little brother the P262 (135w / 8 ohms) at nearly 25% less = US$1300.

Just add a bit perfect streamer such as a WiiM Mini, Sonos Connect (gen1), Raspberry PI or Chromecast Audio, all less that US$100, plus a current DAC with optical in from either Topping, SMSL or Shiit and for well under $2k you will have a SOTA front end. Just add speakers.

There are obviously less expensive systems available but I would propose that you will not get a better quality front end for streaming content from the likes of Tidal and Apple even if you spent another $20k.
 

Rick Sykora

Major Contributor
Forum Donor
Joined
Jan 14, 2020
Messages
3,513
Likes
7,017
Location
Stow, Ohio USA
Good point.

Yes, but AFAIK, RF filtering is part of the EVAL1 and most other buffers based on it.

So, good but not a major differentiator for this amp.
 
Last edited:

peng

Master Contributor
Forum Donor
Joined
May 12, 2019
Messages
5,614
Likes
5,167
It is not realistic to read a power value from a chart that shows amplifier is clipping and the distortion value is changing 100 fold within a few watts.


The 500W reading is not 1% as it is listed on the post nor on the spreadsheet though is it?

I think it can be realistic as it depends how you look at it. To me, I would not buy an amp that clips at 100 W if I need 50 W for my use at the maximum spl I would ever listen to at my mlp. Instead, I would go for a 200 W amp for my 50 W worst case scenario requirement.

So in my opinion, 100 W at clipping is the final limit of an amp that should not be approached under the absolute worst condition when in use.

To some users, they may prefer an amp that distortions would rise more gradually pass the clipping point whether that point is defined as 0.1%, 1%..., or the point where distortion starts to rise sharply. They may feel that in that case they can use their amps right to the limit or even exceed it slightly, without hearing distortions.
 
Top Bottom