I went from Class D to Luxman A/AB - And most of what you think is wrong

Hi everyone,

As most of you know, I’m a fan of Class D. I have lived with ICEPower 250AS based amps for a couple of years. Before that I lived with a pair of Parasound A21s (for HT) and now I’m listening to a Luxman 507ux.

I have some thoughts after long term listening:
  • The tropes of Class D having particularly bad, noticeable Class D qualities are all wrong and have been for years.
  • No one has ever heard my Class D amps and gone: "Oh, wow, Class D, that’s why I hate it."
  • The Luxman is a better amp than my ICEPower modules, which are already pretty old.

I found the Class D a touch warm, powerful, noise free. Blindfolded I cannot tell them apart from the Parasound A21s which are completely linear, and run a touch warm due to high Class A operation, and VERY similar in power output.

The Luxman 507 beats them both, but no amp stands out as nasty sounding or lacking in the ability to be musical and involving.

What the Luxman 507 does better is in the midrange and ends of the spectrum. It is less dark, sweeter in the midrange, and sounds more powerful, almost "louder" in the sense of having more treble and bass. It IS a better amplifier than I had before. Imaging is about the same.

There was one significant operational difference, which others have confirmed. I don't know why this is true, but the Class D amps needed 2-4 days to warm up. The Luxman needs no time at all. I have no rational, engineering explanation for this. After leaving the ICEPower amps off for a weekend, they sounded pretty low fi. Took 2 days to come back. I can come home after work and turn the Luxman on and it sounds great from the first moment.

Please keep this in mind when evaluating.



Showing 13 responses by atmasphere

This amp opened my eyes what digital can do, I'm not a digital snob anymore. I'm now wanting to look at more current digital amps to see what there all about, mine is quiet but you do hear a little noise if you put your ear close to speaker.
Just for the record, the fact that class D and the word 'digital' both start with a 'D' is coincidence (which is confusing to the market)- its D because A, B and C were already taken. Class D amps are not digital although they are switching. They are an entirely analog process.
And for the most part in the nCore/ICEpower/Pascal/TI world this is true, though recent amplifiers from Technics blur the line further, with an A/D stage, DSP processing and unknown amounts of analog feedback at the outputs.
Yes, that does add to the confusion. But even with an digital source built into the amp, its still an analog process.
Calling it a digital amplifier is a lot closer to the truth than calling an analog amp.
Not if you talk to a designer of a class D amp :)

Switching is a bit different from digital. Digital has bits with which meaning (a sound signal) is imposed- how many bits are on or off and in what combination carries the meaning. The switching of a class D amp does not. Instead, it switches according to its encoding scheme. Gain is not represented by how many bits are on or off, gain is instead produced by the encoding scheme, for example in a Pulse Width Modulation scheme the gain is essentially the difference of the triangle wave vs the input signal. Thus it is possible to vary the gain of the amp by changing the amplitude of the triangle wave. This is fundamentally different from digital!
For example, I’ve read that GaN transistors are capable of switching so fast they literally eliminate dead-time and any resulting distortion caused by it.

If a designer believes that, its simply because the encoding scheme and related parts are slow enough that the circuit may not need any dedicated deadtime circuitry since its literally built in due to those slower parts.  Or- the designer added enough heatsinks so that the GaN devices stay cool even though there is shoot-through current.

Your need to go on ad nauseum without actually addressing the points I made is actually kind of like watching a fetish in public.
What point did I not address?

And this, folks, is why this hobby is in a death spiral. Stupidly expensive gear with adherents that can "hear" all sorts of artifacts and distortions which they focus on to the detriment of enjoying the music. Had a very interesting chat with my 30 year old software engineering son. He questioned why anyone would spend the kind of money high end companies charge to listen to music. His generation looks upon this hobby as a total waste of time and resources, and that is the future of audio. Go to RMAF or any of the other big shows and take note of what you see. For the most part, a bunch of 60 something white guys with more money (sometimes) than sense. Class A, Class A/B, Class D, Class G, ad nauseum.
Not sure I agree with all of this. Go to Munich and watch for the young families with strollers and under 30 years men and women. Its quite different in Europe and Asia!

Regarding distortions- anyone can hear distortion since the ear uses higher ordered harmonics to sense sound pressure. Traditional solid state amps are well-known for making such harmonics and being quite audible as harshness and brightness. This is why tubes still exist!

Kids these days buy LPs- hence the revival of the medium- kids keep the music industry alive! Whether they are conscious of it or not, they are hearing something in the LPs that they like as opposed to digital. You don't have to know anything technical to understand that.

I see people of all ages at RMAF and AXPONA but admit that younger people are a minority. What works for kids is the same as it was for us old people- that its relevant to our lives. When they hear music reproduction that sucks they just stick with their earbuds. When the reproduction is good, then they dream of the day when they can have that sound at home. The latter is what happened to a lot of us older audiophiles. The shows are there so the public (and hopefully kids) can feel the dream.

Regarding class D, the carrot on the stick is that a class D output section does not have the same sort of artifact that is imposed by the output section of a traditional solid state amp. The distortion made by the amp is thus caused by the encoding scheme, whatever is needed to drive the encoder (and some comparitors have pretty low input impedances, so this is not a trivial matter) and how much deadtime is needed to keep the output transistors happy. As switching transistors get faster (the current embodiment of speed being GaN devices) the deadtime needed is reduced and with it the distortion of the amp.

I am confident I am not alone in saying GaN is well beyond my budget,
@tweak1   FWIW these days GaN devices do not cost any more than regular MOSFET switching power transistors and in some cases are cheaper. Their implementation in a class D circuit does not cost any more either. Armed with this knowledge it might be interesting to ask a manufacturer of a GaN-based class D amp why their amp is so much more expensive.
Ralph, not meant to be insulting, but couldn't you sell your products for a lot less? Nothing there that's cutting edge, though I do understand that highest quality parts cost significantly more

And wouldn't the GaN input buffers need to be a lot more exotic= $$$?
All of our products are built by hand, and to that end set a pretty high bar in terms of construction quality. Despite that we are expecting our first class D entry to be less expensive than most of our amps are right now.
Uh, George, GaN amps still need deadtime. It *reduces* the need but does not eliminate it.
Deadtime is not a major source of distortion in a class D amp; you can double the amount used above the amount EPC recommends and not be able to hear it. OTOH the accuracy of the encoding scheme plays a much larger role in how much distortion is made.

The reason Technics is switching so fast is not because they don't need any deadtime. Its been done to reduce the residual waveform to a very low level and open up the bandwidth of the amp.

BTW 'switching noise' really isn't a thing unless the switching frequency is really low, in which case the amp won't be suitable for audio. IOW you're simply not able to hear it or any knock-on effects- plain and simple. 

Which it's artifacts are, even the non technical can see it in any Class-D's 1 or 10khz square wave from on the amps outputs, and in the phase shift the filter produces down to 4khz in many cases.
You can't see switching noise on any frequency a class D amp can amplify. All that can be seen is a sine wave known as the 'residual'. You've been told this before. That sine wave is at the switching frequency of the amplifier; if you are saying you can hear 250KHz or 1.2MHz sine wave good luck trying convince anyone that you aren't a nutbag.

Regarding phase shift, filters in most class amps are set to 60-80KHz and are usually 12dB per octave; phase shift thus derived is less that you would see with a 6dB slope (with a 6dB slope phase shift can be seen down to about 1/10th the cutoff frequency). So the 4KHz thing in the statement above is just plain false. You might see something at 20KHz though, but it would be slight even on an older class D amp.

I know your [sic] bringing out your own Class-D amps just like Ricevs is also using off the shelf boards (maybe slightly modded) that's maybe why the negatives toward the GaN Technology, that now even Texas Instruments are behind.
We have our own circuit and are not using anyone's boards. We filed for our patent over a year ago and expect the patent soon.

twisting things to suite your self, I never ever said that you can hear these frequencies, I’ve always stated they have an effect and the "output filters used" in all Class-D amp to remove them have an audible effect, because they cause problems!!, and if the "output filters" are shifted up higher because the switching frequency is higher, then problems become minimized.
Hm. Your prior statements do not agree with the statement above. Here's an example- take a look:
GaN technology will finally get rid of these flaws of dead time and **switching noise**.
(emphasis added)
-what then, is 'switching noise', if not the residual??

In addition, as I pointed out before, most class D amplifiers use a 12dB slope in the filter. Now if you're up on your filter theory, you know that the phase shift artifact of a 12dB filter does not extend down to a 10th the cutoff frequency- that is a 6dB/octave filter that does that.

As for you getting abusive
I don't recall doing that and can't find any example of such in my prior posts. As you know, being abusive to another member violates forum rules. So I am quite careful in all my posts to avoid that! If you are referring to my prior post wherein this text occurs:
if you are saying you can hear 250KHz or 1.2MHz sine wave good luck trying convince anyone that you aren't a nutbag.
-that text is not abusive since I'm pretty sure that you are not in fact trying to convince anyone that you can hear to 250KHz or higher. If such is not the case, please let me know as that will be a bit of a surprise!

The only interest you Ricevs and Mike Mivera have coincidently in degrading anything to do with Class-D GaN Technology, is because it poses a big threat to all your yet to be released non GaN class-D amps. Which you all so delicately push here on these threads, except for Mike who got what seems to be a permanent sabbatical away from Agon for pushing too hard.
Sorry, don't know what you're talking about. I can't recall nor can you point to a post of mine where I've 'degraded anything to do with Class-D GaN Technology' as you put it. FWIW we've been working with GaN FETs for quite some time. IMO it makes no sense not to.
@georgehifi If this comment is directed at me, could you explain how I was twisting your words? Are you saying that surface mount boards are only done on machines and can't be done by hand? How am I abusing you? Have you tried reporting my post to moderation? I think you will find that the post stands because I've made no personal attack, although I have debunked your statements which is allowable under forum rules.
@kijanki Thanks for your elucidation of some of what's involved with surface mount. I think a lot of people think that since we make tube gear by hand that somehow that means we don't have an education or something- or don't keep up with the times. We're using 603 components right now- as you point out, easy to do by hand. And probably too for production- depending entirely on volume of sales! On this basis, this statement:
it's impossible to create a "retail product" complete Class-D board, using all smd components, making it by hand soldering.
You say it can be done, but that would only be a 1 off prototype and not the production version.
-is simply false. Now at some point we may go to 402s just to reduce inductances and resulting ringing and radiation. But for now the 603s have worked out alright. If 'sales' translates to 100s of shipments, then its likely we will farm out the work.
it's impossible to create a "retail product" complete Class-D board, using all smd components, and making it by hand soldering
There appears to be a sleight of hand going on here, compare the above quote above to this one that appeared earlier:
To do that you need to buy either ready made boards then do simple mod and call them your own, if not and you want to do your own GaN Technology boards, you need the knowledge to design it from scratch, and you need some serious automated wave soldering machines to do all those smd parts that are involved, you can’t do it by hand. 
When comparing these two, we see that George had to back off from the idea that 'to do all those smd parts that are involved, you can’t do it by hand'; that's morphed into you just can't create a 'retail product' by hand soldering. Both statements are false. The soldering industry has all sorts of products designed for hand soldering work- and one of my employees worked doing production smd work all by hand.
Except for maybe a prototype, good luck hand soldering for retail production a Class-D that’s all SMD components like this cheap $34 Class-D stereo.
We're not planning to build a $34.00 amplifier and to be clear a $34.00 amp isn't going to have GaN FET output devices- such an amp would be all entirely on one chip. Its realistic to discount that comment as irrelevant.