ATI AT522NC Stereo Amplifier Review

[audimus]

I am not against more measurements but why did this become a useful measurement and particularly for Class D? Are there amps known to send bad things outside of audio bandwidth? Didn’t know it was a thing.

By the same token it also make sense to make sure an amp rejects such high frequencies at the input.

 

[JohnYang1997]

Curious:

View attachment 39547

Maybe not specific to this amp, but...

I'm used to seeing power harmonics, and test tone harmonics, but not used to seeing spikes in other frequencies.

They are low enough, but where are they coming from? Is it a Class D thing?

Compare to

View attachment 39548

I think it's smps.

 

[amirm]

I am not against more measurements but why did this become a useful measurement and particularly for Class D?

There has been concern on and off about ultrasonic energy from these amplifiers. The previous measurement were at 5 watt showing relatively high amplitude of ultrasonics relative to main tone. In this latest meaurement, I am showing max power which now shows the ultrasonics to be some 70 dB lower, indicating that the switching noise doesn't scale with signal proportionally.

 

[pma]

Could you post a 10kHz square response at some 5Vp-p?

Could you post a spectrum if input signal is 440kHz and/or 900kHz sine 10mV amplitude?

 

[audimus]

Would a non Class D amp get such a free pass for not measuring close to advertised power specs? Or that many spurious harmonics even if total distortion number is lower? Is there a “Class D sound” from those harmonics?

We typically roast manufacturers for not measuring to their advertised specs. Don’t we? Or do Class Ds get a soft spot in the House of SINAD because of what they measure for that number?

At 20 Hz though, distortion shoots way up and we are able to only produce 22 watts before THD+N of 0.1% is hit. Fortunately there is not much music energy there so practical audible concern is low.

And this is a forum that feels cold at DACs that reach SINADs in the hundred, because we expect better in numbers whether that affects audibility or not.

FYI: There is a typo there. Should be 20khz. Also Watts misspelt in the picture above it.

Is this kind power drop at high frequencies typical for all classes of amps? Yes they may shutdown for heat protection with a full band sweep but say if one measures just say 10k to 20k?

Kind of makes the FCC requirement of average power across full bandwidth beg for the fallacy of averages. Perhaps they should be specifying deviation as well so manufacturers don’t game it with just limited band full power.

 

[amirm]

Would a non Class D amp get such a free pass for not measuring close to advertised power specs?

The advertised specs have fine print of certain amount of THD+N. That is not my standard. I pick the knee in the curve. I also explained in the review that the knew is so sharp that more points can help extract a few more watts there.

So in summary, no, all amps I measured and judged the same as me. And my method doesn't match what the industry/marketing department practices. This is the same with class D amps or not. Headphones or Power amps.

 

[pma]

Class D is a fashion now. Such is consumer electronics business. Issues are overlooked.

 

[Kal Rubinson]

The AT522NC is part of a series which starts with 2 channels of class D amplification and goes up to 8 channels in the same enclosure (I assume).

Not all can fit in the same small chassis.

 

[amirm]

Is this kind power drop at high frequencies typical for all classes of amps?

Probably not but note that I have refined the test recently. In the past I left it at default 1% THD. For this review, I dialed it down to 0.1% because the amp would not climb up to 1%. As we get more mileage from this test using new settings, we can start to draw better conclusions.

 

[amirm]

Not all can fit in the same small chassis.

Thanks. Corrected.

 

[blueone]

I have been listening to the testimony in DC, and the above reminded me of what they always say when a witness is confirming their position.

Ok, I’ll bite. What is the significance of a full spectrum to 1mhz sweep with a 40hz signal? Why 40hz and what perspective does this provide? I truly don’t know and I would guess most here don’t. I yield the floor to you...

To add some personal color to what Amir already said, Class D amplifiers introduce ultrasonics into the outputs that have no correlation to the input signals. Hypex, as do all Class D manufacturers I'm aware of, specify that their amplifiers should be measured with AES-17 filters on the measuring equipment, so that ultrasonics are not considered in the measurements for distortion or noise. Some speakers have tweeters that have frequency responses well beyond the knee in the AES-17 filter curve, which makes me (and others) curious about what the broadband output really looks like, unfiltered. Amir is the first reviewer I've personally seen who is studying the ultrasonics in the outputs, so I'm applauding that. I think it's interesting, and I'd like to develop an opinion independently of what the amplifier manufacturers tell me I should do.

What do I think about these measurements? Superficially, noise that's only 24db below 5W could raise some eyebrows. 5W is a very interesting power level to people with moderate to low-efficiency speakers in large listening rooms. On the other hand, if I did the math correctly, that's less than two one-hundredths of a watt of noise power. And the full power chart, which is still unlikely to happen even at 40Hz for more than a few seconds, perhaps on an organ music track, is showing the noise to be 68db below the fundamental, which seems very unlikely to harm any speakers.

Would I buy this ATI amp, based on Amir's measurements? Probably not. Nor a Hypex variant with a switching power supply. I totally get the efficiency benefits for Class D amps, but I like the trade-offs in the Benchmark Media amp better. That's just me.

 

[March Audio]

Can you tell me which tweeters have response well beyond the 100kHz flat bandwidth of Amir's input filter?

 

[amirm]

To be clear, the FFT tests are performed without a filter. The rest are.

 

[audimus]

The advertised specs have fine print of certain amount of THD+N. That is not my standard. I pick the knee in the curve. I also explained in the review that the knew is so sharp that more points can help extract a few more watts there.

So in summary, no, all amps I measured and judged the same as me. And my method doesn't match what the industry/marketing department practices. This is the same with class D amps or not. Headphones or Power amps.

I saw that but the difference was more than a few watts.

But fair enough, though it would be useful to explicitly compliment manufacturers that quote specs that is more like what you measure (honest ratings) like say the Outlaw vs those that follow the industry marketing specs that diverge a lot from the knee measurement. If you believe these reviews influence the industry, that is.

On a related note, I don’t see a single spec sheet on the Nord UK site. They don’t publish them or do they just point to the Purifi or NC reference specs elsewhere? The Nord three review does not state any spec power ratings either for the unit itself.
Never mind, found it.

 

[Matias]

Class D is a fashion now. Such is consumer electronics business. Issues are overlooked.

They measure better than the majority of amplifiers out there, are very cheap for their state of the art in band performance, small, light, powerful, and are available from multiple vendors in various configurations and formats. I call this a good fashion! Customers win.

Issues are always raised by armchair analysts, while people who actually own and use them are satisfied. I am on my 3rd model of class D power amplifier having owned 3 different class AB integrateds before.

 

[LTig]

Probably not but note that I have refined the test recently. In the past I left it at default 1% THD. For this review, I dialed it down to 0.1% because the amp would not climb up to 1%.

How can this be? Is there a limiter inside?

 

[blueone]

Can you tell me which tweeters have response well beyond the 100kHz flat bandwidth of Amir's input filter?

Doesn't Hypex specify the AES-17 20KHz brick wall filter for measurements? Or have I misunderstood the notations in their data sheets?

For example, the Scanspeak Illuminator Be tweeter has an on-axis frequency response that extends to at least 40KHz, which is beyond the knee of the AES-17 brick wall filter. Other Be tweeters have similar frequency response capabilities, which is one reason I was curious.

https://www.madisoundspeakerstore.c...or-d3004/6040-00-beryllium-dome-tweeter-each/

 

[amirm]

How can this be? Is there a limiter inside?

Not in the direct manner. The analyzer attempts to find 1% THD, not any value way below or above that. It will hunt back and forth with input voltage to try to achieve this and if after so many tries it can't get there, it will quit. It is a pain as it can quit 90% into the test, requiring messing with input level limits to try again and again. This is how I killed my Sony Receiver. With some amps, no matter how much I raise the input voltage, the analyzer is unable to find this magic point. With two trials, I could not get it to find the right setting for 1% THD, but did get there for 0.1% THD.

 

[Audiocrusader]

Let's not forget this chart. Who here is under 20 and can hear up to 16k? And who here is older and can only hear till 8K?

 

[LTig]

Not in the direct manner. The analyzer attempts to find 1% THD, not any value way below or above that. It will hunt back and forth with input voltage to try to achieve this and if after so many tries it can't get there, it will quit.

Understood. Stupid software though.

I once wrote an algorithm to control the gain of a receiver such that its signal optimally feeds an ADC. If the gain was set too high so that the ADC clipped (digitized signal reaches 0dBFS) the algorithm tried a new gain in the middle between the current setting and the last one which did not clip the ADC. If it did not find a perfect gain after a few few tries it would just use the highest working gain found so far.