ATI AT522NC Stereo Amplifier Review

[restorer-john]

As I just explained, those measurements are academic anyway.

We are trying to evaluate their performance.

The capacitive/overload recovery test was widely regarded as an indicator of an amplifier's potential stability into real world speaker loads. You, yourself have expressed interest in a load box, or designing one, so we know you understand it's a missing part of your tests. Remember, your friends at Audio Precision say this:

"Reactive testing, where the load includes inductive and capacitive components, is not as commonly performed. Real world speakers, however, often have significant inductance and capacitance. Neglecting to test an amplifier with reactive loads means we can only speculate about its performance in actual use. An amplifier that has good resistive power output ratings, but poor drive into low impedances or reactive loads, may produce distortion well before it reaches rated output."

You have just found an amplifier that produces a tiny fraction of it's rated power at 20KHz before its distortion skyrocketed. And that was just into a fixed resistive load! We know the output stages of those amplifiers have a huge amount (approx 2uF or ~4R@20Khz...) of parallel connected capacitors to ground. And we know they'd likely have issues with high power bandwidth testing due to that.

You've got the equipment, the knowledge and the DUT, and yet there's a curious push-back to investigate your own findings. Me, I'd be all over it like a cheap suit.

In addition, you need differential probes or dual channel to avoid shorting out the output of the amps that have floating negative. And if you want a load attached at the same time, I would need yet another adapter cable to make it so.

These modules do not have a floating negative. Have a look at the schematics.

 

[amirm]

You have just found an amplifier that produces a tiny fraction of it's rated power at 20KHz before its distortion skyrocketed.

No I didn't. That test has 90 Khz bandwidth so it is pulling in noise shaping. THD vs frequency tests always have issues like this.

 

[RayDunzl]

Square Wave...

I'd like to see the AP's harmonic display on one of those as far as it will go, just for fun.

This (10Hz square) was one posted here earlier, with lesser precision, and less bandwidth:

(click and double-click to expand the image)

I estimated 1,150 harmonics displayed, much to my surprise, both in them being generated, and in the ability of the software to see them all at once.

S'prise me mo'.

 

[RayDunzl]

Hey what’s with the “ce” logo?

https://en.wikipedia.org/wiki/CE_marking

and

https://en.wikipedia.org/wiki/CE_marking#China_Export

https://starfishmedical.com/blog/conformite-europeenne-mark-vs-a-china-export-mark/

https://www.europarl.europa.eu/sides/getAllAnswers.do?reference=P-2007-5938&language=EN

Just another example of IP theft, I suppose.

 

[restorer-john]

They just stuck it there, the China exports logo- makes no sense.

No "using local and imported parts"

https://www.linkedin.com/pulse/open...3023430549504&trk=public_profile_article_view

It's the same as the cars we designed, engineered and manufactured in Australia that used some US made V8 engines. Thing was, Australians tore off the Holden badges and put Chevy badges on their cars.

 

[Ron Texas]

@restorer-john I think you have a point. I wonder how well some of these amplifiers with stunningly low measured distortion would compare when hooked up to a speaker with it's complex passive crossover and voice coils oscillating in a magnetic field.

 

[restorer-john]

No I didn't. That test has 90 Khz bandwidth so it is pulling in noise shaping. THD vs frequency tests always have issues like this.

You said this:

The noise doesn't scale with power though. Plotting the same at max power with source frequency of 40 Hz gives us this:

Does the noise scale with frequency at high power? What happens? Let's see 20KHz at full rated power on the same plot.

The capacitive issue also needs to be investigated. There's plenty of audiophile speakers with capacitive loads at mid to high frequencies (Quad ESLs etc).

 

[amirm]

You said this:

And since then played with it more and it seems to be ultrasonic noise issue. Power increased sharply when I lowered the bandwidth to 45 kHz:

And I also mentioned to not run with that graph until there is more data on it.

 

[pma]

Hypex 20kHz output waveform:

 

[RichB]

Hypex 20kHz output waveform:

View attachment 39876

It looks to be .25 volt swings on the waveform that should not be there.
How should I inerpret this?

- Rich

 

[pma]

It looks to be .25 volt swings on the waveform that should not be there.
How should I inerpret this?

- Rich

It is the carrier frequency, about 0.9Vp-p, about 440kHz.

 

[Bear123]

I'm not remotely close to pulling the trigger on an amp yet, but from the research I've done, it seems like the Nord Hypex NC500 based 3 channel amp(since I want a 3 channel) is the best all around value if one factors in cost, power, and measured performance. Their 3 channel is under $1700 shipped with 12 volt trigger and a discount code. Its quite a bit cheaper than the Benchmark and offers quite a bit more power although it's measured inaudible distortion is not as low as the Benchmarks inaudible distortion. It also seems to measure very similar to ATI's version, just with a lot more power and a much lower cost, even compared to their 2 channel version. Anything I'm missing?

 

[Matias]

I'm not remotely close to pulling the trigger on an amp yet, but from the research I've done, it seems like the Nord Hypex NC500 based 3 channel amp(since I want a 3 channel) is the best all around value if one factors in cost, power, and measured performance. Their 3 channel is under $1700 shipped with 12 volt trigger and a discount code. Its quite a bit cheaper than the Benchmark and offers quite a bit more power although it's measured inaudible distortion is not as low as the Benchmarks inaudible distortion. It also seems to measure very similar to ATI's version, just with a lot more power and a much lower cost, even compared to their 2 channel version. Anything I'm missing?

If you need 3 channels, a pretty good value is Nord's amplifier based on NC252MP.
3x250W @ 4Ω
https://www.nordacoustics.co.uk/product-page/nord-one-mp-nc250-2-5-channels

 

[kaka89]

It is the carrier frequency, about 0.9Vp-p, about 440kHz.

Do we have this in class A or AB amp?

 

[pma]

No.

 

[DonH56]

Some have a little, e.g. those using SMPS supplies (sometimes coupled with switching/tracking rails, class G or H), but I have not seen any with as high a level at the output as class D amplifiers. But remember it is far above the audio passband and at a low enough level to not bother a tweeter.

 

[pma]

Here you go:
View attachment 39663

OK. Looking at this individual measurement, I do not like the result. Oscillations on top and bottom of the squares may indicate to instability or to output filter resonances. Longer impulses (lower repetition frequency) would help. To measure rise time precisely (difficult to read from your plot), another measurement with shorter time base would be needed. Please let me show some examples:

as for step response rise time

As for long square (20Hz), that perfectly reveals LF response, LF roll-off (top and bottom should be straight)

 

[miero]

100Hz square wave sampled at 44100Hz (and then 8x oversampled) goes from min. to max. level in appox. 40us.

 

[pma]

100Hz square wave sampled at 44100Hz (and then 8x oversampled) goes from min. to max. level in appox. 40us.

I do not understand your point. To read properly rise time of the amplifier output, you need fast sampling, at least 20MHz. On the other hand, "long" squares, like 100Hz, reveal decay of impulse top (bottom) due to low frequency limit of the frequency response (high pass filter response). "Middle" repetition frequency of squares, like 1kHz, may indicate to settling time of ringing. I do not understand why the square wave should be sampled at 44.1kHz.
https://cnx.org/contents/i34SwKCi@1/AE-Lecture4-PartD4-Time-Domain-Response-of-High-Pass-Filter

Square wave for amplifier testing should be taken from analog generator output. Output of the amplifier to be measured by reasonably fast oscilloscope, BW > 20MHz.

 

[miero]

OK, I just wanted to note that the fastest response that is required for listening to audio (without HF noise considered) is not so high.

And if we need to test it with "faster" squares then it is not needed for (CD quality) audio, but for detecting other issues which will be revealed in such test.