Benchmark...first watt....ABX...facepalm!

[amirm]

They sure are doing something odd there. For one, i don't know how they got a relative graphs to 1 watt in AP. But even so, the noise level should be rising but it is not.

That's the problem I have with stuff like this where the details are sorely lacking to figure out what the data means.

 

[DonH56]

Technically the noise is not rising but rather the signal level is decreasing so noise becomes a larger fraction and THD+N rises due to the "N" part.

Sounds like they are weighting out the noise.

 

[tomelex]

In the first link, down at the bottom of the page is a graph. It's labeled:
"THD+N vs Power - Relative to 1 W"

I wonder what the "Relative to 1 W" part means?
I also don't recall ever seeing a graph where the noise doesn't increase as the power decreases.

***************************************
Here is a very nice DIY chip amp kit graph.

You might be overthinking this. Don't take offense, as I did this for me as much as for an answer to your question as it is a good point that I had just glossed over since I intuitively knew what they were on about.

DBR is db relative to something, db are just ratios, so they need something to "be attached to". In their graph, the y axis is relative to 1W of power (note their little circles on their colored lines relative to 1watt on the x axis)measured probably on a resistor. I don't like math, but db are just an easier way to represent really big or small numbers really.
thus:

Odb we will reference to 1W thus dbr , dbrelative is to our defined 1 watt
+100 dbr (watt) is then 10,000,000,000 more of watts
-100 dbr (watt) is then 10,000,000,000 less watts

So, simply they are showing that their amp has less THD+N than the other guys using their selected thing "to be attached to", ie 1watt across some load as their 1 unit. Their dbr, dbrelative to 1 watt might (agree with Don here as well) also have used an "A" weighting filter so it a little bit better corresponds to how our ear roughly evaluates loudness (loudness is our internal amplitude reference from not hearing anything hardly to a bomb going off next to us)

Clearly, Benchmark need to better explain their stuff, use more pictures etc, so folks can really grasp what they are saying, they should say their THD+N is 150 thousand times lower or whatever it is than their competitor, that is easier to understand and sounds really impressive...

 

[Cosmik]

... there is also AG and AH.. These amps (don't know if anyone makes one) are class A up to several watts or even tens of watts, then if the input signal is higher their power supplies output higher voltage during that time, so if done right, you get "relative" efficiency and also capability to reach huge peak output when needed. So you get advantages of Class D and class A both.

Presumably such an amp would need to "see into the future" to know when a large signal peak was coming. Does this mean they can only be used with a digital buffer and some latency?

 

[Don Hills]

No, it's simpler than that. There are 2 schemes that I an aware of. With the one described by Tomelex, called class G or class H operation, there are 2 or 3 power rails with increasing voltages, and a "totem pole" of output devices fed from the rails. As the output voltage swing approaches one of the rails, the next higher device turns on. There's some illustrations on the amplifier Wikipedia page:
https://en.wikipedia.org/wiki/Amplifier#Classes_G_and_H
With the other scheme, there is a push-pull class A amplifier with a low voltage power supply. This keeps the power dissipation low. The centre of its power supply is in turn connected to a class B amplifier output. Both amplifiers have the audio signal at their input. As the output voltage swings towards to the class A stage power rails, the class B stage "moves" the low voltage supply, keeping the class A stage within its linear range. For example, the class A stage might have +- 6 volt rails. If the output voltage has to swing to 12 volts, the class B amplifier lifts the centre point of the low voltage supply to 12 volts. The class A stage only has to swing far enough to compensate for the the crossover distortion and nonlinearities of the class B stage. As far as I know, this was unique to one manufacturer (Yamaha I think). I'll have to refer to my (paper) files...

 

[Fitzcaraldo215]

...

However, I should note that even class A needs to stabilize for some time to alleviate this effect (tube or transistor) but once stabilized, no matter if high and low cycling the final output is still near perfect. As the class A is first turned on, the tube or transistor general heating is taking place, and thus even though the bias is solid, the heating of the transistor parameters or the tube elements requires a settling in time.

...

My subjective, anecdotal listening impressions in my Class A days support this. I had a pair of mid-90's Krell Audio Standard-2 Class A monoblocks for about 15 years. It seemed to me consistently that they required 20-30 minutes after coming out of Standby status to sound decent.

That was one of several reasons I finally replaced them with a Spectron Musician III Mk 2 Class D, which I also thought sounded better in general. Each Krell had also blown up in a big puff of smelly smoke, requiring costly repairs. I could also clearly see a drop in my monthly electric bill after the change. I am still happy with the Spectron 6 years later. And, even after 15 years, the Krells had fetched enough used to pay for the Spectron. The same dealer who had sold them to me saw my Audiogon ad and bought them back, unaware at first that I was the same guy who bought them originally.

Incidentally, that Krell and later ones had the tiered bias thresholds described by DonH, three levels as I recall. There were indicator lights on the front to let you know which output tier was currently being used. That was in addition to a big, beautiful output power meter, which was rather useless because it had a linear rather than a logarithmic scale.

 

[DonH56]

IIRC class G uses switched ("stepped") rails, often just a few steps. Class H uses a tracking power supply, at least that is what I have always known as the distinction. Depending on the source the actual base amplifier can be just about anything, though class D wasn't really defined when the class G amd H designations came about. I think, working from a fuzzy memory...

The implementations I have seen use the input signal in a feedforward manner to increase the rail voltage on demand, depending upon the latency of the circuit. A purely analog approach. That in turn implies minimal charge storage (no big bank of filter caps) and/or very high current capability to provide the prerequisite power supply slew rate. This is not as bad as it seems, as the largest signals are typically the lowest in frequency, so the power supply's rate of change can be pretty slow and still provide substantial benefits in efficiency without significant distortion. One of the tests I used to do "back in the day" was a pulsed 20 kHz sine wave that stepped from say 1 W to 100 W. Most of the class G/H amps would not switch in a cycle and some went unstable. For that matter, some regular class A and AB amps were not very happy with such a test... Pretty unrealistic as a real-world test but was entertaining at times.

Class A circuits were often designed with minimal feedback and relatively simple bias circuits so needed a few minutes to warm up. The time I measured, again ages ago so don't bother asking for proof, ranged from a few seconds to several minutes with a large factor being the amount of bias, how stable the bias circuit, and thermal management (primarily heat sinking). I seem to recall one that ran very hot and had very simple biasing that would wander around even after the normal warm-up period. The actual shift in performance was fairly small but it was noted for taking an hour or so to warm up. I had a tube amp that also seemed to take a while, and in that case the bias point also shifted significantly with temperature due to both tubes in the power supply and a string of zeners that had a pretty strong tempco. A later tube amp also had tubes and zeners but in a regulated supply circuit for the low-level stages so exhibited less shift even though the output stage bias wandered around.

When I started doing high-speed moderately high resolution (for the speed) ADC and DAC design I developed a program to analyze the thermal changes within the circuit that led to nonlinearity with signal. It was interesting (I still have my notes, somewhere) and the effect was bigger than we anticipated. It was interesting in that I was able to devise some test circuits that verified my analysis. It was less interesting since I did not have a real good way to counter the mechanism without burning more power. One of the solutions used a tracking input stage akin to class H power amps (OK, distantly akin). It worked, but was costly in power and area.

 

[Blumlein 88]

No, it's simpler than that. There are 2 schemes that I an aware of. With the one described by Tomelex, called class G or class H operation, there are 2 or 3 power rails with increasing voltages, and a "totem pole" of output devices fed from the rails. As the output voltage swing approaches one of the rails, the next higher device turns on. There's some illustrations on the amplifier Wikipedia page:
https://en.wikipedia.org/wiki/Amplifier#Classes_G_and_H
With the other scheme, there is a push-pull class A amplifier with a low voltage power supply. This keeps the power dissipation low. The centre of its power supply is in turn connected to a class B amplifier output. Both amplifiers have the audio signal at their input. As the output voltage swings towards to the class A stage power rails, the class B stage "moves" the low voltage supply, keeping the class A stage within its linear range. For example, the class A stage might have +- 6 volt rails. If the output voltage has to swing to 12 volts, the class B amplifier lifts the centre point of the low voltage supply to 12 volts. The class A stage only has to swing far enough to compensate for the the crossover distortion and nonlinearities of the class B stage. As far as I know, this was unique to one manufacturer (Yamaha I think). I'll have to refer to my (paper) files...

I believe that a very early Pass amp also worked like the Yamaha.

 

[Cosmik]

Ultimately, it sounds equivalent to any scheme that aims to combine different output stages e.g. current dumping. What is notionally the power supply becomes just another 'modulator', and feedback mops up the details.

Very interesting to hear about ways to upset amplifiers - isn't this the audio holy grail? Objective proof that amplifiers that "measure the same" are not, in fact, the same if you use unconventional methods to test them. I know very little about amplifier design, but I have always been slightly perturbed by the absolute certainty that amplifier designers seem to have about the characteristics of standard circuit blocks and their schemes' immunity to speaker loads and unusual signals.

 

[fas42]

Very interesting to hear about ways to upset amplifiers - isn't this the audio holy grail? Objective proof that amplifiers that "measure the same" are not, in fact, the same if you use unconventional methods to test them. I know very little about amplifier design, but I have always been slightly perturbed by the absolute certainty that amplifier designers seem to have about the characteristics of standard circuit blocks and their schemes' immunity to speaker loads and unusual signals.

What's even more amusing is that it's trivially easy to model real world conditions against a schematic modeled in Spice, say, and see the amplifier fail to behave correctly, that is, not as good as conventional testing would imply. So typical amplifiers even theoretically are a long way from being "bullet proof" - but cheap and easy implementation is the name of the game - it's only audio, after all ... .

 

[Nightlord]

The Benchmark amp is one of the few ones out there while really interests me to have a listen to... drawbacks being that it's a bit too low on power and that it's not been reviewed by the people I want things reviewed.

( Regarding class G/H - a very good example of that was the NAD 208. Only real issue with that was build quality, a lot of them have needed service over the years. )

 

[amirm]

Regarding class G/H - a very good example of that was the NAD 208. Only real issue with that was build quality, a lot of them have needed service over the years.

First, warm welcome to the forum Nightlord. It has been a while!

On this, at my company we were thinking about carrying them but were strongly warned not to do so due to poor reliability of NAD products. I hear the latest stuff may be better though.

 

[Purité Audio]

The Benchmark amp is one of the few ones out there while really interests me to have a listen to... drawbacks being that it's a bit too low on power and that it's not been reviewed by the people I want things reviewed.

( Regarding class G/H - a very good example of that was the NAD 208. Only real issue with that was build quality, a lot of them have needed service over the years. )

The AHB2 is a really nice amp, it has replaced some Mola-Mola 'Kalugas' here, quiet,compact, efficient and powerful, ideal really.
Keith

 

[Nightlord]

First, warm welcome to the forum Nightlord. It has been a while!

On this, at my company we were thinking about carrying them but were strongly warned not to do so due to poor reliability of NAD products. I hear the latest stuff may be better though.

Yeah, the 208 sticks out over the years in 2 regards.... Very close on being the ultimate amp and the worst reliability. Very unusual and clever design below the lid, I'm told by people who really know (=professional amp designer).

I'd say it is a very good brand to carry in its price range. Might shame the more expensive stuff now and then too.

NAD s300 was the same product as Gryphon Tabu if I remember correctly. Drastic price tag difference, though.

Bit off topic now....

 

[oivavoi]

Out of curiosity, in case I ever come across a NAD 208: what kind of issues have come up? Can they stop working even with normal use?

 

[Nightlord]

Out of curiosity, in case I ever come across a NAD 208: what kind of issues have come up? Can they stop working even with normal use?

Very good question. It never happened to me and I'm not sure I've ever read of when they break, if it's on power-on or otherwise. I've never heard of a failure that damaged other gear, though.

I would myself think that if it's still alive today, it's probably a good specimen or already repaired.
I'd go for it if I were you... It's capable of handling dynamik peaks above the kilowatt. Here in Sweden they go for $700-1000, but elsewhere where the test results of it are less known, it can be a bargain.

 

[DonH56]

Google shows a plethora of threads on the NAD 208 power amp and its woes (and proponents, natch). One of the stores I worked at way back when carried NAD and there were a few years we saw high failure rates, seemingly related to high operating temp.

 

[Sal1950]

Google shows a plethora of threads on the NAD 208 power amp and its woes (and proponents, natch). One of the stores I worked at way back when carried NAD and there were a few years we saw high failure rates, seemingly related to high operating temp.

I tend to lump NAD of the time period with Adcom. Both offered some highly reliable, best bang for the buck gear. And then they had some stuff that was not either designed or built to the same level. I've owned various Nelson Pass designed Adcom amps (GFA-5x5 and 5x00) for over 20 years and they are awesome components. But I've also owned 2 GSA series surround pre-pro's that couldn't keep working for more than 6 months at a time. The first series I bought was replaced with a later model after the first failed the third time within 2 years. That new one failed twice within a year. I was so pissed I literally opened the listening room window and threw it 2 flights down to it's death. DONE, OVER. I blame some of the QC issues with Adcoms moving of the assembly from Japan to China sometime in the later 90s, but for sure there's much more to the story than that.

 

[Nightlord]

Just repeating it one more time... there's NAD in general... and there's NAD 208. It's a very special beast and way above the rest in performance. I know people who've been thinking of approaching NAD about buying the rights to it and start making them again - with improvements to the weak spots made, of course.

It's just one of those rare products that stick out...

 

[Blumlein 88]

Just repeating it one more time... there's NAD in general... and there's NAD 208. It's a very special beast and way above the rest in performance. I know people who've been thinking of approaching NAD about buying the rights to it and start making them again - with improvements to the weak spots made, of course.

It's just one of those rare products that stick out...

I respectfully disagree. Sunfire amps used a similar idea. They were better amps. NAD 208 isn't that special. Not even as special as some of the newer NAD class D offerings. Back in those days a Conrad-Johnson MOSFET of equal power and greater current output without the problems was a much better amp. Those will have resale values for years to come. Made like a brick shitehouse. I beat one badly for years on my horridly inefficient Soundlabs and all it did was put out nice music without complaint or issue. Dead quiet amps too. Even it doesn't match my current class D Wyred4Sound amps. They sound almost like a monstrously powerful Spectral.