NEW GaN Amplifier : Dual ACM8816 GaN amplifier chips + AK4118AEQ [YJCO D816]

[torgeirs]

By the way the infineon evm link uses a driver that is also used for power suppplies and motors.
But for audio the loop of analog feedback makes it all better than the digital feedback (ADC in the loop) often used for motor drive or power supply

 

[fpitas]

I posted a YouTube video of it in this thread where you can hear this amplifier, even compared to a similarly expensive analog Class A/B amplifier.

Seriously, and for the sake of lurking onlookers, a YouTube video is completely inappropriate as a comparison method when you're talking modern low-distortion electronics.

 

[torgeirs]

Also remember now that sawtooth signal for making the pwm need to be really high bandwith. So in that EVM they use a high quality clock and analog high speed comparator.
For motors and powersupply they use the MCU clock and a counter with limited speed. So the driver gets a much less clean signal than for audio. Still about 60-70 THD with digital if done right

 

[john61ct]

There is no ADC involved here, amirite?

> THD: 0.02% | SNR: 108dB

How can SINAD be so poor if these #s are true?

What are the power output specs again, can it drive 2Ω ?

 

[Michael Kohler]

The output low pass filter is usually an LC filter (a coil inductor plus a capacitor). It gives a 12 dB/octave attenuation. An LC filter has very little damping, and the actual damping relies on the resistance of the load (the loudspeaker) to dissipate the energy. There the lower the load resistance, the higher the energy dissipation and higher damping. That's why you usually see a higher response peak at the filter cutoff frequency for an 8 Ω load than for a 4 Ω load. This problem of load dependent frequency response can be effectively mitigated using post-filter feedback, as in the cases of Purifi, Hypex, Axign, Icepower, many of the newer TI chip based amplifiers with PFFB, etc.

Below is a block diagram of an Axign based amplifier. Since the Axign operates in the digital domain, all the digital signal processing rules and techniques apply. The brick wall filter is from the digital processing of the signal.

I'm going on some traveling starting today for the next few weeks. So my posting will be sporadic.
View attachment 558480

Thanks, NTK!
In that context, I remember a YouTube video with Bruno Putzeyes, where he talked about feedback in audio amplifiers and how high feedback rates aren't a bad thing. On the contrary.

Later, I also read Axign informational documents, in which the manufacturer stated that significantly higher feedback rates can be achieved with their digital feedback technology compared to analog feedback. Back then, I didn’t yet know that you can’t just increase feedback arbitrarily because an amplifier can become unstable and possibly destroy itself.

Now that I’ve looked at your picture of "MPS," I remember it again, because In the text, the manufacturer claims again that their digital feedback is consistent and programmable. But there must be a downside, right? Because the amplifiers with an Axign module measure worse than an high Quality analog one, especially the ones from Bruno, for example the Purifi ones.

 

[Roland68]

Thanks, NTK!
In that context, I remember a YouTube video with Bruno Putzeyes, where he talked about feedback in audio amplifiers and how high feedback rates aren't a bad thing. On the contrary.

Later, I also read Axign informational documents, in which the manufacturer stated that significantly higher feedback rates can be achieved with their digital feedback technology compared to analog feedback. Back then, I didn’t yet know that you can’t just increase feedback arbitrarily because an amplifier can become unstable and possibly destroy itself.

Now that I’ve looked at your picture of "MPS," I remember it again, because In the text, the manufacturer claims again that their digital feedback is consistent and programmable. But there must be a downside, right? Because the amplifiers with an Axign module measure worse than an high Quality analog one, especially the ones from Bruno, for example the Purifi ones.

What exactly do you mean by "worse"?

 

[Michael Kohler]

And then we can ask ourselves. Is it maybe better to put a 24 bit audio bandwith DAC in front and use a 100 mhz opamp with -140 dB THD for the feedback?
Doing it digitally requires far less snalog components of cource. So that is a really nice benefit

What I don’t understand: You once wrote that in audio technology, a goal is to reach about 120 dB SNR. Why 120 and not, say, 130 or 100 dB? I always thought the CD standard, which doesn’t even reach 100 dB, is more than enough in reality, or real life. Is it better to have 120 dB, or is it (always?) better to have 130Db or more?

 

[Michael Kohler]

What exactly do you mean by "worse"?

Guten Tag Roland.
By 'worse,' I mean that the technically measured values for distortion and stuff, posted here on ASR by Amir, are worse than those of a modern Purifi amp. Not much worse, just a little, but on 'paper,' so to speak.

 

[Roland68]

There is no ADC involved here, amirite?

> THD: 0.02% | SNR: 108dB

How can SINAD be so poor if these #s are true?

What are the power output specs again, can it drive 2Ω ?

No, no ADC.

I have my concerns regarding this chip and its measured performance.
The design reminds me a lot of the basic Infineon chips—and unfortunately, so do all the measurements I’ve seen so far.
There are various Chinese USB/BT/DSP/DAC ICs that are very cheap and used in all sorts of devices, such as soundbars, Bluetooth speakers, headphones, etc.

Devices tested on ASR using such combo ICs have shown very poor measurements, dragging down the performance of all other components and sections.
A case in point is the review of the SMSL A50 Pro; the Infineon amplifier ICs used in it performed significantly better in other SMSL amplifiers—almost as if they were completely different components.

I’m very curious to see real, independent measurements of the YJCO D816 and the 3E Audio GaN Micro. I have a bad feeling about them, but I’d be happy to be pleasantly surprised.

 

[GaNMaster]

I’m very curious to see real, independent measurements of the YJCO D816

I will probably take measurements next week.

 

[Michael Kohler]

Great. I'm so excited. So I want to go ahead and write now about how it sounds.
So, for everyone who doesn't care for subjective sound descriptions, please stop reading now.

The quick answer: It sounds amazing, really.

The long answer: With my setup, speakers (Martin Logan Ethos), it almost reaches the quality of my favorite amplifier, the sabaj a30a.
I can hear a significant difference between the 48-volt 5-amp power supply and a stronger 48-volt 10-amp one. In my setup, the 10-amp sounds better, just overall. Not by much, but clearly and easy to pick up.
In accuracy, precision, and neutrality, the D818 is a little bit worse than the Sabaj A30a, but: It's an unfair comparison. First, the difference is pretty small, and second, the D816, with a price of around 150 euros including a 10 A power supply, is super cheap.

The first special quality of the D816 is, at least with the 10A power supply, its remarkable power. Big speakers? Not a problem at all. It drives them with ease. Subjectively, even a bit more powerful than my Sabaj a30a, even if not quite that neutral and controlled.
The second strength is its subjective naturalness. It doesn't sound as neutral and "razor-sharp" in detail resolution as, for example, a brand-new Purifi. It can't do that. But that's maybe also a good thing, because it sounds just a tiny bit like a McIntosh or a tube amp: colorful and pleasant.
If you had told HiFi fans 15 or 20 years ago that something like this would exist for such a cheap price, everyone would have laughed at you.

And now I'm really curious about the measurements. What if it measures terribly?

 

[john61ct]

Thanks for that

And now I'm really curious about the measurements. What if it measures terribly?

That adverb is a judgment call as well.

Many will consider say THD: 0.02% / SNR: 108dB / 74 dB SINAD, noise level ≤ 100μVrms

to be "just fine" relative to audibility, having allowed damage to occur to their ears / perception by going clubbing, too many live shows or maybe just affected by age.

Others will only spend their money on specs cleaner even than objectively "golden" ears cannot detect.

And I have no idea if there is even close to a consensus here on whereabouts that boundary is.

Well level-matched ABX blind testing with a statistically significant number of trials would be the only way to find out

but then that would be required just to know if you can actually hear the difference between this amp and your Sabaj

much less whether or not you really prefer the latter due strictly to acoustics, as opposed to other factors, biases you are not even aware of.

None if which is to make my gratitude for your feedback any less sincere

 

[MaxBuck]

If you want, you can listen to the sound of this device. I posted a YouTube video of it in this thread where you can hear this amplifier

If you can "hear" an amplifier, it's not a good amplifier.

 

[Michael Kohler]

Thanks for that

That adverb is a judgment call as well.

Many will consider say THD: 0.02% / SNR: 108dB / 74 dB SINAD, noise level ≤ 100μVrms

to be "just fine" relative to audibility, having allowed damage to occur to their ears / perception by going clubbing, too many live shows or maybe just affected by age.

Others will only spend their money on specs cleaner even than objectively "golden" ears cannot detect.

And I have no idea if there is even close to a consensus here on whereabouts that boundary is.

Well level-matched ABX blind testing with a statistically significant number of trials would be the only way to find out

but then that would be required just to know if you can actually hear the difference between this amp and your Sabaj

much less whether or not you really prefer the latter due strictly to acoustics, as opposed to other factors, biases you are not even aware of.

None if which is to make my gratitude for your feedback any less sincere

In this post you write that the D816 has an SNR of only 74 dB. I had read it and thought, how does he know that? The manufacturer doesn’t state it directly. Then I googled and found out that anyone can easily convert it. The problem is that 74 dB is actually a really poor value, if it’s true. What I can say is that the amplifier actually does hiss, the noise is always the same regardless of the volume setting, but you only hear it with your ear directly on the speaker, so for me that’s with my ear right up to the panel of my Logan Ethos. As soon as I move my ear about 3 centimeters back, the noise is no longer audible.

 

[john61ct]

74 dB is actually a really poor value

The penny drops! That was stated on the same day this thread was posted.

If your room is always noisy, you sit far away from speakers and/or they are insensitive

then YOU may not care about the difference (or even hear it, need a proper test to know that)

Lots of tube lovers put up with far worse.

But, not many of them hang out here, since those sort of specs are what we care about, not subjective claims based "trusting my ears"

 

[john61ct]

Maker claims SNR = 108dB

but Google says this is Idle Noise only

> The chip maker (ACME) designed the ACM8816 GaN chip to be exceptionally quiet when no music is playing. If you aren't playing anything, the background hiss is pushed down to an inaudible -108 dB. This means you will not hear static coming from your speakers

Whereas the (claimed) THD: 0.02% applies when signal is flowing, so ~74dB

The claims for GaN tech bringing an advantage to SQ rather than just heat efficiency, IF TRUE might mean our current measurement technology might not show the full picture.

All of which is speculation.

 

[Michael Kohler]

Maker claims SNR = 108dB

but Google says this is Idle Noise only

> The chip maker (ACME) designed the ACM8816 GaN chip to be exceptionally quiet when no music is playing. If you aren't playing anything, the background hiss is pushed down to an inaudible -108 dB. This means you will not hear static coming from your speakers

Whereas the (claimed) THD: 0.02% applies when signal is flowing, so ~74dB

The claims for GaN tech bringing an advantage to SQ rather than just heat efficiency, IF TRUE might mean our current measurement technology might not show the full picture.

All of which is speculation.

You're right. I just tried it. With my ear right on the panel, no noise, zero. Doesn't matter if the D816 is connected to the speaker or not. I also know now what I heard buzzing this afternoon: the blood in my ears. No joke. It was a self-deception.
Even though it doesn't matter in reality, the 74 dB bother me. Or rather, my vanity. Are there even other modern amplifiers that are also specified at 74 dB? Do you know any?

The truth is, I really like the little amp. The sound is balanced and pleasant, yet VERY powerful. I know some people here don't believe me, but it's true—the 10A power supply makes a difference in terms of subjective power. If you don't believe me, just buy the 5A, but I strongly recommend going for the 10A. Financially, it doesn’t matter either, the difference is only 10 or 20 euros. For the double power! Which the amp could actually reproduce as well. Every Purifi or Hypex power amp has even more power, and no one says anything about that.

 

[john61ct]

Of course the extra input power creates higher SPL if your speakers are thirsty and you turn it up!

And even a tiny SPL increase results in people rating SQ better, that's why precise level matching in blind tests is critical.

And it is also "possible" with this novel Full Digital Amplifier tech there is something going on with increased current availability, besides just the "moar headroom" with analog amps.

Just don't bother with "it sounds better to me it's true!" blathering. "I know it might be my imagination, but my subjective perception is X, Y and Z" would at least show you're learning...

 

[Svend P]

I know some people here don't believe me, but it's true—the 10A power supply makes a difference in terms of subjective power.

It is not like we think you are lying. You do just not test the right way, so you are fooling yourself.

I might as well say I could tell the difference between red and green M&Ms. If you pointed out that I looked at them when I was eating them, my reply would be "yeah but the difference is so huge it doesn't matter if I look or not. The red taste red-ish and the green taste green-ish". No one would believe that either.

You know you are connecting a more powerful power supply, and your brain goes into overdrive. "Yeah, more power, moaaaar... faster, harder (Scooter)" and that is the result you get.

 

[john61ct]

Even a basic SPL measurement would show if there is that difference