A Simple 10W Direct-Coupled Class-A Power Amplifier

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by mediatechnology »

I was reading about Benchmark's new power amp and found this statement:
Low Gain – The AHB2 has a low gain setting that optimizes the gain structure of professional monitoring systems. Maximum rated output is reached at an input signal level of 22 dBu. This places the upstream equipment in an ideal operating range to maximize the SNR of the monitoring chain. Most power amplifiers have far too much gain, and this degrades noise performance of the overall system. The AHB2 has an ultra low-noise input amplifier with two gain boost settings that can be enabled to allow direct interfacing with Hi-Fi components that usually operate at significantly lower signal output levels.
Siau, "A Radical Approach to Audio Power Amplification" http://benchmarkmedia.com/blogs/news/14 ... lification

This has been on my radar recently both with the Class-A headphone amp and my 10W Class-A.
Most power amps have way too much gain for the way we use them.
It seems insane to have 35 dB of voltage gain in the power amp and then throw most of that gain away in the amp's input level control.
Sure there are times when you need the voltage gain for PA applications but in most home or small studio settings you don't.
One of the reasons the Dual Class-A headphone amp is so quiet is that it's unity gain.

I've been told that one of the reasons most manufacturers use such high gain post-input level, rather than to vary PA gain via feedback is to be able to spec high slew rates.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by JR. »

mediatechnology wrote:I was reading about Benchmark's new power amp and found this statement:
Low Gain – The AHB2 has a low gain setting that optimizes the gain structure of professional monitoring systems. Maximum rated output is reached at an input signal level of 22 dBu. This places the upstream equipment in an ideal operating range to maximize the SNR of the monitoring chain. Most power amplifiers have far too much gain, and this degrades noise performance of the overall system. The AHB2 has an ultra low-noise input amplifier with two gain boost settings that can be enabled to allow direct interfacing with Hi-Fi components that usually operate at significantly lower signal output levels.
Siau, "A Radical Approach to Audio Power Amplification" http://benchmarkmedia.com/blogs/news/14 ... lification
The best audio myths contain a sliver of truth and this one does. Yes indeed operating the power amplifier at modest gain will reduce the contribution from it's own input noise, but seriously how significant is that input noise? (rhetorical query it isn't significant IMO.)
This has been on my radar recently both with the Class-A headphone amp and my 10W Class-A.
Most power amps have way too much gain for the way we use them.
It seems insane to have 35 dB of voltage gain in the power amp and then throw most of that gain away in the amp's input level control.
Sure there are times when you need the voltage gain for PA applications but in most home or small studio settings you don't.
One of the reasons the Dual Class-A headphone amp is so quiet is that it's unity gain.
There are several more marketing than actual application reasons for elevated gain in power amps.

#1- uninformed customers often confuse high gain with higher power. In uncontrolled informal listening tests the louder amp will usually win, even if it isn't capable of making more power when properly driven than the softer amp.

#2- the actual voltage gain "needed" in a power amp is enough gain to make full power from the signal available from the worst case probable preceding stage. Modern gear with single ended outputs can make 8-10V, so gain needed depends on the amp power with more voltage gain needed to saturate a higher power amp. Some amp manufacturers try to standardize voltage gain within an amp family to simplify setting up multi-way (active crossover) speaker systems. Some amplifiers provide sensitivity switches to accommodate different popular gain structures. (If you want to confuse amp customers ask them what amp sensitivity means?) :lol:

#3- PA amplifiers that provide clip-limiting (like Peavey's DDT) need more voltage gain than just driving the output to full scale before clipping the previous stage to take full advantage of the clip limiter for some "free" limiting. Peavey's DDT provides something like 20 dB of gain reduction (or more) so for full use of that limiting you need to drive the amp to 20 dB past normal output clipping, without clipping the input or previous active stages.
I've been told that one of the reasons most manufacturers use such high gain post-input level, rather than to vary PA gain via feedback is to be able to spec high slew rates.
There is a relationship between minimum closed loop gain and the stability compensation capacitor required so arguably higher gain requires less compensation capacitance and results in faster slew rates all else equal. My understanding is that slew rate does not dominate the design decision. A very difficult consideration for commercial amp design is to be unconditionally stable no mater what load the customers hang off the amp. I have seen some very unfriendly load lines from esoteric passive speaker crossovers, where some speaker engineer tweaks the crossover for best response ignoring what that means to the amplifier. Another issue is capacitance from funny speaker wire and inductance from driving transformer isolated constant voltage systems. You might be amused by the worst case speaker loads the amp engineers used at Peavey to prove an amp was always stable (I don't recall exactly but tens of uF, etc). High closed loop gains increases the phase margin for better stability. Slew rate has not really been a design issue since the '80s when they had the "slew rate wars" to impress the uninformed more-is-better customer (followed by the damping factor wars, etc).

An amp that breaks into oscillation because of a funny load can catch the speaker on fire (I've heard of that) and/or self destruct from mutual conduction (both up and down devices turned on at same time) common in some mosfet amps or burning up the output trap (LC to ground) that is not sized to handle the power from full scale sine waves at hundreds of kHz. :oops:

JR

PS: for one off designs do what works for you, large scale amp makers try to be all things to all people and with amp performance very much improved it is harder and harder to differentiate your new model from the pack. I give Benchmark an A for effort... nice story.
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by mediatechnology »

John - What do you make of the feed-forward error correction?
In reading it I'm reminded of the Quad current dumping topology.
I never seriously looked at what Quad did but what my understanding was is that it had a small amp to handle the zero crossover region and higher power dumpers in what I suppose was a class-B stage.
I realize that the Benchmark also has class-H elements, but doesn't the feed-forward aspect sound like Quad's current-dumping technique?
I know that its marketing material but the feed-forward aspect doesn't sound new.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by JR. »

Please don't make me read it closely.. from a quick pass it seems like an collection of popular buzz words-concepts cobbled together with some rather dubious claims thrown in (like large reservoir caps are bad :lol: :lol: ).

I give them an A for marketing effort. About the design I'd have to see a schematic (that I don't want to see either) but I expect it is serviceable since they have a decent reputation. At this point they already sound good so they just need to convince the customers that they are special somehow.

FWIW Class H amps actually introduce perturbations a little like crossover distortion when the PS rails get stepped. More attractive for efficiency than Hi-fidelity while they can be made quite good with fast switches etc.

JR

[edit I wasted some time looking for THX amplifier patents. I found one application that sounded a lot like the PMA70+ amp I did at peavey back in the '80s with the boosted rails (100+W peak, 35W continuous).. they had another patents for alternate variations on NFB in an amp. We never patented my old Peavey amp, so doubt a patent search would turn anything up. [/edit]
Cancel the "cancel culture", do not support mob hatred.
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by ricardo »

mediatechnology wrote:Siau, "A Radical Approach to Audio Power Amplification" http://benchmarkmedia.com/blogs/news/14 ... lification
The really radical part of this design and one which will give it a distinct advantage in DBLTs is its regulated PSU.

I conducted extensive DBLTs on the prototypes of a 50W commercial HiFi amp in the late 1980's. Most of these came up with NULL results as expected. The unexpected result was between 2 toroid PSUs of the same VA rating (ie same cost. 200VA IIRC) .. one giving the correct 2x50W @ 8R for the design .. and the other wound to give 2x45W @ 8R.

My true golden pinnae panel had a small but significant preference for the 'smaller' amp. It turned out that the amps were clipping more often then expected and the slightly lower rails which were also slightly stiffer cos less resistance meant that on overload, the yucky 100Hz sawtooth that is imposed on the signal was less.

The AHB2 is medium power (100W 8R) so I expect its regulated PSU to give it an advantage over similar power but unregulated PSU amps. I don't expect this to hold above the 200W 8R level.

You might argue that clipping a 50W amp is misuse but our own investigations suggest this is far more common among home users than most designers believe. You can check if this applies to you on http://www.diyaudio.com/forums/multi-wa ... -need.html

Alas the Marketing Dept insisted on the higher voltage toroid so they could put 2x50W 8R on the brochure :oops:

Today, a well behaved regulated SMPS might be cost effective wrt to a large conventional amp supply.
_____________________________

AFAIK, I am the only one who ever conducted DBLTs on the above but there's another important non-trivial result from DBLTs on amps. This time, other people have also done independent tests and come up with the same answer.

The really "good for better sound" feature is severe bandwidth-limitation .. brickwall at 20kHz or even more drastic. My tests were with vinyl and mastertapes. But there have been DBLTs this Millenium (Jan Didden of LInear Audio was involved in one IIRC) that show this STILL applies with EVIL Digital .. a surprising result for me as EVIL digital is usually already brickwalled. The results are pretty well unanimous for both Jurassic & modern tests of this.

Of course this is even more unlikely to be accepted by the Marketing Dept. :o
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by JR. »

Happy next year already Ricardo..
ricardo wrote:
mediatechnology wrote:Siau, "A Radical Approach to Audio Power Amplification" http://benchmarkmedia.com/blogs/news/14 ... lification
The really radical part of this design and one which will give it a distinct advantage in DBLTs is its regulated PSU.
Huh? Did you perform a blind listening test on this benchmark?
I conducted extensive DBLTs on the prototypes of a 50W commercial HiFi amp in the late 1980's. Most of these came up with NULL results as expected. The unexpected result was between 2 toroid PSUs of the same VA rating (ie same cost. 200VA IIRC) .. one giving the correct 2x50W @ 8R for the design .. and the other wound to give 2x45W @ 8R.

My true golden pinnae panel had a small but significant preference for the 'smaller' amp. It turned out that the amps were clipping more often then expected and the slightly lower rails which were also slightly stiffer cos less resistance meant that on overload, the yucky 100Hz sawtooth that is imposed on the signal was less.
As I've shared I've been exposed to too many single blind tests where the pukes (not golden ears) preferred clipping to a pretty benign clip-limiter (Peavey's DDT which by that time had decades of tweaking for transparency). The listening panel "preferred" the sound of letting the amp clip, ripple modulation and all. They thought the kick drum had more SNAP... :lol: :lol:
The AHB2 is medium power (100W 8R) so I expect its regulated PSU to give it an advantage over similar power but unregulated PSU amps. I don't expect this to hold above the 200W 8R level.
class H ? While I haven't seen a schematic, and I'm not sure why such a modest power amp needs class H. Typically class G/H are used at above 1000W power points.
You might argue that clipping a 50W amp is misuse but our own investigations suggest this is far more common among home users than most designers believe. You can check if this applies to you on http://www.diyaudio.com/forums/multi-wa ... -need.html
This designer believes... A great deal of design effort goes into managing saturation, from fast recovery device clamps, freedom from introducing DC components when clipped, and even benign sounding limiters, so you can't clip. .
Alas the Marketing Dept insisted on the higher voltage toroid so they could put 2x50W 8R on the brochure :oops:
Probably the the paying customers too....
Today, a well behaved regulated SMPS might be cost effective wrt to a large conventional amp supply.
Unless you roll PFC into the regulated switcher it's six of one half dozen another as you still have PS caps charged at the mains frequency. Crest did some big dog amps with regulated PS but I don't know that they were ever considered by the GE community.
_____________________________

AFAIK, I am the only one who ever conducted DBLTs on the above but there's another important non-trivial result from DBLTs on amps. This time, other people have also done independent tests and come up with the same answer.

The really "good for better sound" feature is severe bandwidth-limitation .. brickwall at 20kHz or even more drastic. My tests were with vinyl and mastertapes. But there have been DBLTs this Millenium (Jan Didden of LInear Audio was involved in one IIRC) that show this STILL applies with EVIL Digital .. a surprising result for me as EVIL digital is usually already brickwalled. The results are pretty well unanimous for both Jurassic & modern tests of this.

Of course this is even more unlikely to be accepted by the Marketing Dept. :o
Lets hope 2015 is better than 2014,,, while it's always good to still be vertical and drinking cold beer....

HNY

JR
Cancel the "cancel culture", do not support mob hatred.
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by ricardo »

JR. wrote:
ricardo wrote:
mediatechnology wrote:Siau, "A Radical Approach to Audio Power Amplification" http://benchmarkmedia.com/blogs/news/14 ... lification
The really radical part of this design and one which will give it a distinct advantage in DBLTs is its regulated PSU.
Huh? Did you perform a blind listening test on this benchmark?
No I haven't JR. But I've done enough DBLTs to bet money that it would sound "cleaner, faster bla bla" than any 100W 8R competitor with conventional PSU.
As I've shared I've been exposed to too many single blind tests where the pukes (not golden ears) preferred clipping to a pretty benign clip-limiter (Peavey's DDT which by that time had decades of tweaking for transparency). The listening panel "preferred" the sound of letting the amp clip, ripple modulation and all. They thought the kick drum had more SNAP... :lol: :lol:
JR, the results of a DBLT tell you what those who aren't deaf prefer. (A good DBLT will have measures in place to weed out the deaf ones too)

The Man in the Street's preference is surprisingly consistent with true golden pinnae. It's the deaf Golden Pinnae who hear "chalk & cheese" that are (totally?) inconsistent.

I can well believe that lots of people (including true golden pinnae) prefer a certain amount of clipping on certain types of music (?!) It's the same phenomena that makes the DJ turn up the music until ALL the lights on the amp flash in time with the music. They equate loud with distortion. :mrgreen:

Of course on other types of music, even a very small amount of clipping is detectable and objectionable to most people.

In http://www.aes.org/e-lib/browse.cfm?elib=3793, we found music (???) where clipping 50% of the time was undetectable :o

When I do DBLTs, the victim ALWAYS chooses the music & playback level.

The type of clean Peavey clipper you describe would appear to make the sound softer cos it 'turns down the volume control'. I've not done DBLTs on this type.

I have on crude diode soft clippers which tend not to do that but change the character of the clipping. Here, the 'soft clipping' is preferred.
Unless you roll PFC into the regulated switcher it's six of one half dozen another as you still have PS caps charged at the mains frequency.
I would expect PFC in any amp SMPS today. Cordell & one or two others talk about this on another forum.
_________________________________

Also a reminder that a Class A amp would benefit most from a regulated supply cos
  • ripple is huge
  • amp is smaller so will be overloaded more often
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by JR. »

ricardo wrote:
JR. wrote: Huh? Did you perform a blind listening test on this benchmark?
No I haven't JR. But I've done enough DBLTs to bet money that it would sound "cleaner, faster bla bla" than any 100W 8R competitor with conventional PSU.
That is a significant logical leap that PSU alone could predict such a result.
JR, the results of a DBLT tell you what those who aren't deaf prefer. (A good DBLT will have measures in place to weed out the deaf ones too)

The Man in the Street's preference is surprisingly consistent with true golden pinnae. It's the deaf Golden Pinnae who hear "chalk & cheese" that are (totally?) inconsistent.

I can well believe that lots of people (including true golden pinnae) prefer a certain amount of clipping on certain types of music (?!) It's the same phenomena that makes the DJ turn up the music until ALL the lights on the amp flash in time with the music. They equate loud with distortion. :mrgreen:

Of course on other types of music, even a very small amount of clipping is detectable and objectionable to most people.

In http://www.aes.org/e-lib/browse.cfm?elib=3793, we found music (???) where clipping 50% of the time was undetectable :o

When I do DBLTs, the victim ALWAYS chooses the music & playback level.

The type of clean Peavey clipper you describe would appear to make the sound softer cos it 'turns down the volume control'. I've not done DBLTs on this type.
neither have I but single blind tests, using typical listeners revealed a preference for louder over cleaner. PS DDT is not a clipper but fast attack-fast release limiter using OTA for gain reduction.Result of this limiting is less average output power
I have on crude diode soft clippers which tend not to do that but change the character of the clipping. Here, the 'soft clipping' is preferred.
I did some bench work in the late '70s looking at alternate ways to extend headroom by modifying the clipping waveform that was VERY unsuccessful. It sounded much worse than normal clipping (to my unblind single puppy listening test). My design philosophy for recording consoles was to sample at multiple stages of the audio chain so clipping could be completely avoided.

Severe clipping in sound reinforcement is caused by the all to common newbie practice of "not bringing enough rig for the gig". An amplifier driven into hard clipping will put out significantly more average power than when driven conservatively which can drive speakers into power compression (voice coil impedance goes up while overheated leading operator to turn up even more). ETC... Effective clip limiting is most useful for inexperienced operators (i.e. Peavey customers).
Unless you roll PFC into the regulated switcher it's six of one half dozen another as you still have PS caps charged at the mains frequency.
I would expect PFC in any amp SMPS today. Cordell & one or two others talk about this on another forum.
and what exactly did they say? PFC is significantly more expensive than a simple SMPS amp, that just chops the mains voltage at HF to take advantage of smaller magnetics. PFC involves significant front end electronics and has pretty much been ignored after the Euro regulators backed off making it mandatory for professional power amps. Note: PFC in Euro zone is mandatory for consumer products but there are cheaper ways to manage power line harmonics at modest power levels.
_________________________________

Also a reminder that a Class A amp would benefit most from a regulated supply cos
  • ripple is huge
In a class A amp, the same PS ripple is present for low power output as high so not masked by loud audio so arguably more audible and/or dirtier at low level when normal amps clean up. IMO a class D amp that has near zero PSR will benefit the "most" from regulation. FWIW Mark Levinson and probably others used regulated rails in their Class A amps.

[*]amp is smaller so will be overloaded more often[/list]
In line with my philosophy of avoiding audible clipping the use of modest power points seems inconsistent for serious listening. The argument that a low power class A amp allowed to clip sounds cleaner than a larger amp that never clips is dubious, but perhaps a personal preference and I don't argue about what other people hear or prefer.

=======

Not to veer back on topic but I think I now understand why they combined class H with their class A amp. The class A output stage biased up for the same class A current at idle as peak power will sit at idle with huge heat dissipation if the power devices see full PS rail voltage too. By using a modulated rail voltage that only makes the PS available when needed will indeed reduce that signature dissipation that class A amps are notorious for. The rationale for pursuing class A design is to eliminate the turn on-off problems associated with crossover distortion. While a class A output stage will not be subject to crossover distortion, the class H introduces another distortion mechanism associated with turning on and off the rail as needed. So instead of a clean DC PS rail, this design sees large switching components. From their FAB I read that they are using fast MOSFET devices for these PS switches (Just like Hitachi did when they introduced class H decades ago). Note: miller capacitance in bipolar power devices will cause some distortion from modulating the PS rails. I saw this in my old boosted rail design. I had to make my modulated PS fast enough to get there in time for a 20kHz tone bust, but not so fast of an edge rate that the miller capacitance feed through was unmanageable. This trade off compromised how much efficiency gain I could actually realize with acceptable distortion levels.

I posted a hypothetical schematic for a class Ab (big A little b) using my 3 device LTP to keep the power devices from ever cutting off. viewtopic.php?f=6&t=679&start=40#p7907
This "should" completely eliminate crossover distortion without the penalty of excessive heat sink dissipation. I might be tempted to run 1A of class A bias to keep the audiophools happy, but I am not suitably motivated to bread board one up, because amps are already suitably clean enough to not be the weakest link... IMO

HNY

JR

PS: I would be surprised if my class Ab hasn't already been done. I think the Japanese designers were looking at something along these lines decades ago. Selling power amps into this market is perhaps more about story than reality. I gave up long ago.
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by mediatechnology »

ricardo wrote:
The really radical part of this design and one which will give it a distinct advantage in DBLTs is its regulated PSU.
Yeah, I really need to order those 18VDC laptop switchers for the 10W project.
My first tests were done with a regulated lab supply.
My true golden pinnae panel had a small but significant preference for the 'smaller' amp. It turned out that the amps were clipping more often then expected and the slightly lower rails which were also slightly stiffer cos less resistance meant that on overload, the yucky 100Hz sawtooth that is imposed on the signal was less. {emphasis added.}
Well that brings us full circle back to the original post and the odd-order compressive distortion at high level (but not at low level) when I ran the dual class-A headphone amp into a speaker load with a measly 85 mA Iq.
The THD measured horribly bad and it sounded good.

I'm glad that I listened before measuring though.
The audible low-level detail was intact and the high-level bass material fatter-sounding.
It sounded good even though the peaks were visually (and measured to be) incredibly distorted from odd-order compressive effects due to running out of current.
Everything sounded a little warmer and wider than usual.
It was a nice effect.

Once I got the power devices and heatsinks connected to increase the Iq current into the +1 amp range the audible warming effect disappeared.
The amp then just sounded more transparent particularly at low levels compared to the class-AB.
Oddly I miss the current-starved one enough to dial back the Iq and listen again.
That may explain why the golden ears liked the "smaller" amp and it's clipping characteristic.
My experience was that it measured bad and sounded good.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: A Simple 10W Direct-Coupled Class-A Power Amplifier

Post by JR. »

I wonder if there's some unintentional loudness contour going on, when over driving a smaller amp (probably no bass boost, but surely some extra HF content)?

Back in the '80s a made a small headphone amp accessory product for AMR/Peavey. To overcome the common complaint about headphone amps not getting loud enough I used 3-channels of amp and drove the sleeve with opposite polarity. Other than that, nothing remarkable about the amp, a pair of complementary TO-220 power transistors biased up class aB (small a big B) driven by decent high speed op amps (TL07x), so it was relatively low output impedance and high output current for a headphone amp.

It sounded good enough driving speakers that you could melt the thermal fuse in the 1A wall wart when played loud, so I limited the average power it could draw from the wall wart by adding a small R in series with PS rectifiers, so it could drive speakers comfortably without breaking (while I never advertised it for that application).

I never spent any time listening to this critically but often used it at trade shows to play CDs through AMR studio monitors in the booth. I don't recall it sounding nasty when clipped and it surely clipped some while driving inefficient studio monitors in the booth during set-up tear down.

JR
Cancel the "cancel culture", do not support mob hatred.
Post Reply