WolfX700 Measurement of Benchmark AHB2 Power AMP

[JohnYang1997]

I said they provide designs, not modules. Hypex or Purifi provide modules implementing their IP. Designs are paper, modules are hardware. It means that the user of the design has to implement, test, debug, ... and doesn't benefit from the volume effect when sharing the module with other. So not simple and expensive.

Rails at +/- 18V => max theoretical power in 8 Ohms is 5W.

https://patents.google.com/patent/US8004355B2/en, chapter "claims". It's a pain in the @ss to read, but this is where everything is.
All the art of patent is finding the little detail that makes a process, an assembly, a method, new. Not related to custom parts. The patent went trough the EPO process, which usually is much more stringent than US or Australia, so I guess there are some novelties. If not, I am surprised nobody tried to invalidate the patent.


You are mixing patents and trademarks. THX will sue companies using EPO patent 2401811 without licensing the same way they will sue the ones using the THX AAA trademark without licensing.
Cost drives 99% of business decisions. For the remaining 1%, @Wes provided a valid hypothesis (exclusive license).

From the extensive measurements I've done, thx888 seems to actually apply the feedforward circuit as the high frequency distortion increases drastically (from -140db to near -120db), it seems to only work best at 1khz. Then/alternatively the power supply is actually what improves the performance just like John from benchmark said the supply is modulated by the input signal hence the common mode distortion is highly reduced for opa564. Then with the feedback correction from opa1612 the distortion is reduced almost to none.

 

[lszomb]

I can think a logic answer to this....if you can sell a power amp with a lot worse measurement at a lot higher price point in the name of hifi-sound, why do they bother to make a better one? Of course not.

The real issue now in the industry is, everyone including those selling thousands dollar power or signal cables, they can sell those products without any issue by promoting these as more musical. We do not have a single standard to proof one sounds better than others.

Thank a lot for the review and a few more metrics tested Wolfman! Also nice to see more images

I got into audio a year and a half ago or so. I saw this amp reviewed, and looked up the company and the product, and realized it was the thing that brought THX to the consumer fore-front. We've now had headphone amps that have made THX a serious name, while also others like Geshelli and JDS proving they can make perhaps an even cleaner amp without their feed-forward error correction approach perhaps (unless they are also doing it themselves in a custom manner).

One thing I simply do not understand is, seeing as how this power amp was being shown off nearly half a decade ago. Why not a single contender has shown interest in surpassing it, or if they have, they've failed in some fashion or another. Is the scaling of the typical power usage that power amplifiers are made for, simply something that doesn't scale very well with the massive fidelity disparage between them, and headphone amps? I understand it plays a factor, but is the factor THIS BIG where for essentially half a decade no one has been able to dethrone this Rolex of the amplifier world?

I have never seen a piece of electronics maintain it's supremecy for this long, especially considering no newer SKU's or revisions of it were ever released(aside from blacked-out front plate).

But more interestingly, how hasn't anyone used the THX tech in any other power amp? We have multiple in other types of amps, heck even in mobile DAP devices now as well. But power amps, it seems just this only one, for some reason.

 

[Tks]

I said they provide designs, not modules. Hypex or Purifi provide modules implementing their IP. Designs are paper, modules are hardware. It means that the user of the design has to implement, test, debug, ... and doesn't benefit from the volume effect when sharing the module with other. So not simple and expensive.

Rails at +/- 18V => max theoretical power in 8 Ohms is 5W.

https://patents.google.com/patent/US8004355B2/en, chapter "claims". It's a pain in the @ss to read, but this is where everything is.
All the art of patent is finding the little detail that makes a process, an assembly, a method, new. Not related to custom parts. The patent went trough the EPO process, which usually is much more stringent than US or Australia, so I guess there are some novelties. If not, I am surprised nobody tried to invalidate the patent.


You are mixing patents and trademarks. THX will sue companies using EPO patent 2401811 without licensing the same way they will sue the ones using the THX AAA trademark without licensing.
Cost drives 99% of business decisions. For the remaining 1%, @Wes provided a valid hypothesis (exclusive license).

You're right I misspoke about modules/designs.

What I meant about their highest spec, was the variant used in the Benchmark Power Amp, not necessarily the designs used by ODM's like Claridy Audio that has seemingly provided the current variants on the market like the Drop 789 and the Monoprice 887 (not sure if SMSL hired them also to create a cut-down version of sorts for their SP200). Perhaps the ABH2 was an exclusive first-ever design that Benchmark and THX worked solely on as a one-off design not offered elsewhere.

As for exclusive license, I don't think that is the case because we don't know what the license is for, how long it lasts, and what limits exist. Obviously Benchmark isn't the only one using THX AAA designs anymore, so when someone talks about a license, what are such terms? The only thing that makes sense to me is that the power-amp design was done in collab with Benchmark, and only they have the design for such. It seemingly simply isn't even on offer even if you wanted it. So I don't even think it has anything to do with license, certainly never seen one for devices last thing long in terms of exclusivity. Wes's hypothesis is valid, though not sound in my book.

I can think a logic answer to this....if you can sell a power amp with a lot worse measurement at a lot higher price point in the name of hifi-sound, why do they bother to make a better one? Of course not.

The real issue now in the industry is, everyone including those selling thousands dollar power or signal cables, they can sell those products without any issue by promoting these as more musical. We do not have a single standard to proof one sounds better than others.

Another thing to attract the most customers possible, it's a marketing checklist sorta thing.

 

[neutralguy]

The multitone level for each frequency is much lower than the single frequency sine.
The sum of all the peaks should be at the same specified level than the single frequency peak, so obviously, each peak is much lower.

How much so depends on the used multitone signal specific Crest factor
For a 32 freq log signal, it's above 12dB for sure.

Let's say the multitone has a crest factor of 12db, which is 9db higher than the single tone's. A 9db reduction in SNR still doesn't account for the 18.5db difference between 108db and 89.5db, so the difference is still due to a rise in distortion.

 

[JohnYang1997]

Two things.
1, The individual tones are truly around -20db comparing to single tone.
2, The multitone comes from the DAC of the APx555 instead of the high performance sine generator. The performance is much worse on the dac. The dac's performance in audio range is actually comparable to adi2dac with some better imd performance.

 

[JRG]

As for feedforward error correction, this is tech older than the Second World War when I looked into it as opposed to negative feedback designs.

That's what I think.
Feedforward's technique exists long before THX or Benchmark.
Sansui used Feedforward extensively in his 80s amplifiers.
Possibly what THX has done is a concrete implementation of that technique along others, but that doesn't mean that it cannot be done another way.
If I'm not mistaken, there are even some commercial ICs that use it internally...
In any case, the difficulty must be in approaching the low levels of distortion of the AHB2

 

[anmpr1]

But more interestingly, how hasn't anyone used the THX tech in any other power amp?

Maybe a proprietary licensing thing? Not sure.

 

[anmpr1]

Feedforward's technique exists long before THX or Benchmark.Sansui used Feedforward extensively in his 80s amplifiers.
2

I owned a Quad 405 amp that used some sort of feed forward. This was late 1970s. I don't know if it was similar to the THX-Benchmark circuit or something completely different. Quad called their circuit a 'current dumping' amp. Not the most poetic of monikers. The 405 was like the AHB2 in that it was a small form-factor device. About twice the size of the old 303, and a quarter the size of something like a Phase Linear. Of course it wasn't that powerful on an absolute basis.

 

[restorer-john]

Rails at +/- 18V => max theoretical power in 8 Ohms is 5W.

Not even close. +/-18V rails will give you +/-12.72V RMS,. minus, let's say, 2V for junction losses, gives us +/-10.7V. Over 8R that is 20 Watts.

 

[Koeitje]

It was because AHB2 had near-perfect measurement results (posted by Amir) that made me determined (albeit a bit difficult) to break through the PowerAmp price ceiling I set myself (about $ 1,000)

Is there any chance you can get a Danafrips Hyperion to measure? It's close to your price ceiling and the specs look good.

 

[boXem]

Not even close. +/-18V rails will give you +/-12.72V RMS,. minus, let's say, 2V for junction losses, gives us +/-10.7V. Over 8R that is 20 Watts.

None of us seems good at this...
10.7*10.7/8=14.3 W RMS

 

[restorer-john]

None of us seems good at this...
10.7*10.7/8=14.3 W RMS

Yes. My maximum was 'theoretical' with no losses. With losses (I allowed 2V per rail) you get 14.3W RMS.

 

[pma]

RMS can't be +/-

 

[restorer-john]

RMS can't be +/-

I am talking about the RMS voltage of the waveform for swing calculation purposes.

 

[neutralguy]

Two things.
1, The individual tones are truly around -20db comparing to single tone.
2, The multitone comes from the DAC of the APx555 instead of the high performance sine generator. The performance is much worse on the dac. The dac's performance in audio range is actually comparable to adi2dac with some better imd performance.

1. As I understand, individual tone levels are only relevant for the FFT graph, whereas the multitone distortion ratio (89.5db) is computed based on average power of distortion versus the multitone, so what's relevant is only the crest factor of the multitone. This often cited paper says that a multitone can be generated to have a crest factor < 6db when the number of tones is a power of two, which 32 is. It would be surprising if the AP implemented something much worse.

2. Thanks for this info. Do we have a loopback measurement of the AP's DAC for comparison?

 

[Rja4000]

This often cited paper says that a multitone can be generated to have a crest factor < 6db when the number of tones is a power of two

As far as I understand it, this is only valid when the frequencies are consecutive multiples of the lowest one.
This is not the case here, since the frequencies are following a logarithmic progression.

On top of it, if I understood correctly, the phases on the AP 32 tones signal are chosen randomly.

 

[JohnYang1997]

1. As I understand, individual tone levels are only relevant for the FFT graph, whereas the multitone distortion ratio (89.5db) is computed based on average power of distortion versus the multitone, so what's relevant is only the crest factor of the multitone. This often cited paper says that a multitone can be generated to have a crest factor < 6db when the number of tones is a power of two, which 32 is. It would be surprising if the AP implemented something much worse.

2. Thanks for this info. Do we have a loopback measurement of the AP's DAC for comparison?

 

[neutralguy]

Thanks @JohnYang1997! Looks like the AP DAC is good enough not to be a factor for the AHB2.

@Rja4000 This is about the limits of what I know about the theory or how AP works. Perhaps @WolfX-700 can comment if the AP shows what the actual multitone levels were, and whether the result is noise limited? It really doesn't look like noise to me because 10W and 100W are similar at 91.4db. Thank you!

 

[Tks]

That's what I think.
Feedforward's technique exists long before THX or Benchmark.
Sansui used Feedforward extensively in his 80s amplifiers.
Possibly what THX has done is a concrete implementation of that technique along others, but that doesn't mean that it cannot be done another way.
If I'm not mistaken, there are even some commercial ICs that use it internally...
In any case, the difficulty must be in approaching the low levels of distortion of the AHB2

I'd like to hear the opinions of folks that have bested the THX offerings without their approach: Tom Christiansen with his HPA-1, Schiit Audio with their Magni Heresy, fellow forum member @Ben1987 with his SONCOZ SGD1 (though only a DAC, the man designed also the Kadas Tone Board, so I'm sure his opinion would be valid here either way), Folks at JDS Labs with their EL Amp II, Fiio with their A5 portable amp that came out back in 2016, Geshelli Labs with their recent knockout Archel2, and finally fellow forum memeber @IVX with his E1DA 9038S and no nonsense approach.

I think many of them may have a good answer as to what it would take to forumlate a power amp that comes close to the far superior performance the ABH2 has that has left so many in the dust for years now.

Or any of our knowledgeable members here, is it simply a complex issue due to the massive power involved that needs to be brought under control, in contrast to headphone amps that don't push much more than a few watts?

 

[wowothanks]

If they do, I will post the measurement results. And will be tracked by means of secret purchases.

狼兄我请教个问题，我想AV跟HIFI一起
然后HIFI这里我想用矩声的element x跟AHB2
正常来说就是AV:AV前级+AHB2+音响
HIFI:element x+AHB2+音响
但是问题在于我怎么样用同样的两个前置音响去接两套系统呢？我是需要选别的型号吗？