Bingo. This is exactly why I came to this forum in the first place. Most here can offer up an honest opinion NOT in direct conflict with their livelihood.
It's just hilarious - they've managed to built quite sophisticated (and most importantly, patented😉) switching mode amplifier and not the switch mode power supply (oversized toroid... as stated)
This is an outcome of digital, or better say, discrete REVOLUTION🙂
Hi,
A "Class D" Amplifier is basically a combination of a Delta Sigma AD Converter running at a few 100KHz sample rate with an equivalent sample rate Delta Sigma "power" DAC, using at best 5 Level Modulators.
By comparison modern ADC's usually run at 128 * times oversampling for single speed, that is at around 6MHz sample rate and they tend to use modulators with 25 to 64 Levels.
In other words a Class D Amplifier has a much lower "raw" resolution as common AD and DA Chips using current technology. The AD and DA chip's generally still are more limited in performance than analogue circuitry.
Without increasing the number of levels dramatically as well as getting the switching frequencies up by a factor 10 - 20 OR finding other ways to overcome the limitations of common switching Amplifiers they must remain generally limited to lo-fi and no-fi applications.
Ciao T
A "Class D" Amplifier is basically a combination of a Delta Sigma AD Converter running at a few 100KHz sample rate with an equivalent sample rate Delta Sigma "power" DAC, using at best 5 Level Modulators.
By comparison modern ADC's usually run at 128 * times oversampling for single speed, that is at around 6MHz sample rate and they tend to use modulators with 25 to 64 Levels.
In other words a Class D Amplifier has a much lower "raw" resolution as common AD and DA Chips using current technology. The AD and DA chip's generally still are more limited in performance than analogue circuitry.
Without increasing the number of levels dramatically as well as getting the switching frequencies up by a factor 10 - 20 OR finding other ways to overcome the limitations of common switching Amplifiers they must remain generally limited to lo-fi and no-fi applications.
Ciao T
Is this opinion is based on listening to class d amps such as those made by Hypex?
Hi,
I am quite familiar with a wide range of both switching and fully digital amplifiers (for the last decade + as it so happens). In isolation or compared to very low grade linear Amp's they sound okay. They do display a wide range of undesirable (to my view) sonic traits.
This is mitigated somewhat as the relative modulation index goes up, that is either very low power class D amplifiers or amplifiers where the modulation index is kept high by adjusting the PSU Voltage to control level - Tact/Lyngdorf and some TI Chipsets.
But even the best switching amp's to come my way fare poorly next to what is essentially just D. Self's Blameless Amp on a Chip, namely the LM3875/LM3886 (never mind more extreme designs).
The underlying problem is that no amount of noiseshaping can get you out of the limits imposed by a sample rate of around 300KHz (tops 1MHz) and a 5-level modulator.
With a bandwidth of 20KHz (that is CD Standard) and a 5-Level modulator (standard in BTL Output Class D Amplifiers, SE Output ones only have 3 Levels and do worse) with 512KHz switching frequency (for ease of calculation) we get a resolution 128 levels or 7 Bit equivalent resolution. or a dynamic range of 42dB (pre-noiseshaping).
To then get 110dB dynamic range (which is pretty shitty for a classic linear amp) in the range below 20KHz we need add 68dB worth of noise shaping, or in other words add fuzzy distortion at -42dB for full scale (or nearly 1% Distortion in traditional parlance).
The worse part about this "fuzzy distortion" is that unlike traditional THD it does not fall with falling levels. Tt is more akin to very bad crossover distortion with a random element added and rises with falling levels. So for a signal with -42dB Level we have equal amounts of fuzzy distortion and signal.
In the end there are ways around (it seems Devialet employs one of them) but they have not found their way into the mainstream.
And why not? Actually, Class D amplification is not THAT much more efficient than optimum Bias Class B (per D. Self's definition). While the theoretical efficiency of Class D is 100%, in reality you rarely get 90% in the amplifier, meanwhile an optimised Class B Amp has 70% efficiency.
So, are 20% gain in efficiency really worth all the problems Class D Amplifiers bring? It seems much of the market votes no, where quality is still of concern over BOM cost. And if we have to add a linear amplifier to handle the < -40dB Stuff together with the switching amp we may actually have a higher BOM cost than making a decent "blameless" AB Amp in the first place.
Ciao T
I am quite familiar with a wide range of both switching and fully digital amplifiers (for the last decade + as it so happens). In isolation or compared to very low grade linear Amp's they sound okay. They do display a wide range of undesirable (to my view) sonic traits.
This is mitigated somewhat as the relative modulation index goes up, that is either very low power class D amplifiers or amplifiers where the modulation index is kept high by adjusting the PSU Voltage to control level - Tact/Lyngdorf and some TI Chipsets.
But even the best switching amp's to come my way fare poorly next to what is essentially just D. Self's Blameless Amp on a Chip, namely the LM3875/LM3886 (never mind more extreme designs).
The underlying problem is that no amount of noiseshaping can get you out of the limits imposed by a sample rate of around 300KHz (tops 1MHz) and a 5-level modulator.
With a bandwidth of 20KHz (that is CD Standard) and a 5-Level modulator (standard in BTL Output Class D Amplifiers, SE Output ones only have 3 Levels and do worse) with 512KHz switching frequency (for ease of calculation) we get a resolution 128 levels or 7 Bit equivalent resolution. or a dynamic range of 42dB (pre-noiseshaping).
To then get 110dB dynamic range (which is pretty shitty for a classic linear amp) in the range below 20KHz we need add 68dB worth of noise shaping, or in other words add fuzzy distortion at -42dB for full scale (or nearly 1% Distortion in traditional parlance).
The worse part about this "fuzzy distortion" is that unlike traditional THD it does not fall with falling levels. Tt is more akin to very bad crossover distortion with a random element added and rises with falling levels. So for a signal with -42dB Level we have equal amounts of fuzzy distortion and signal.
In the end there are ways around (it seems Devialet employs one of them) but they have not found their way into the mainstream.
And why not? Actually, Class D amplification is not THAT much more efficient than optimum Bias Class B (per D. Self's definition). While the theoretical efficiency of Class D is 100%, in reality you rarely get 90% in the amplifier, meanwhile an optimised Class B Amp has 70% efficiency.
So, are 20% gain in efficiency really worth all the problems Class D Amplifiers bring? It seems much of the market votes no, where quality is still of concern over BOM cost. And if we have to add a linear amplifier to handle the < -40dB Stuff together with the switching amp we may actually have a higher BOM cost than making a decent "blameless" AB Amp in the first place.
Ciao T
Part of the reason why class d is only seen in cheap consumer stuff like cheap home cinema receivers and such, while the mainstream hifi still uses class AB and sometimes even a low power class A stage in parallel aka the technics new class A thing.
Read about the design before mocking them, these guys have done their homework on this one😉
Hi,
Well, in terms of efficiency stakes a traditional supply certainly largely kills the only benefit of Class D.
As for the rest, running 4 Amp's interleaved (simple paralleled will not help BTW) does help them to produce a system with more available levels before applying noiseshaping, but even the extra 4 Levels only help a little, to 9 Bit.
So the question on subjective sonics will be strictly down to the noise-shaping implementation and how well this manages to hide itself from human auditory system.
Objectively it is easy enough to pass an MLS Stimulus through both the Class D Amplifier and a similar Linear Amplifier, making sure to sample with sufficient sample rate and bandwidth (which is non-trivial, but doable) and compare the two output results.
Ciao T
Well, in terms of efficiency stakes a traditional supply certainly largely kills the only benefit of Class D.
As for the rest, running 4 Amp's interleaved (simple paralleled will not help BTW) does help them to produce a system with more available levels before applying noiseshaping, but even the extra 4 Levels only help a little, to 9 Bit.
So the question on subjective sonics will be strictly down to the noise-shaping implementation and how well this manages to hide itself from human auditory system.
Objectively it is easy enough to pass an MLS Stimulus through both the Class D Amplifier and a similar Linear Amplifier, making sure to sample with sufficient sample rate and bandwidth (which is non-trivial, but doable) and compare the two output results.
Ciao T
Yep, I would agree with this. Going to a scale small enough and all systems are digital. Oddly enough my amps were designed by de Paravacinis erstwhile business partner, Ian McCarthy. And frankly they are the best amps I've heard.
Also every amp I have ever directly compared (ab'd) sounded different.
However the differences are small and mostly only audible in direct comparisons. IME it is largely differences in damping factor which are audible.
You need to do a little more research before making these patently false statements...There are class d amps being sold by the likes of Bryston, Bel Canto, Levinson, PS Audio, Halcro, Rowland, and the list goes on.....
Last edited:
You also need to do some homework. Not all class d amps operate this way. Read up on the UCD as designed by Bruno Putzeys. This amp has proven itself time and again in comparison to very expensive class a and ab designs.
Hi,
I am quite familiar with it. While the technological approach differs minimally, in the end we are dealing with a looped DS Modulator.
The same can be said of 1930's Tube Amplifiers.
I have heard them on many occasions. I so far decline to use them. A gainclone, assuming the limited power available from one suffices is a better choice to my ears (and is pure Class B BTW).
But to each his own. I'll take a nice tube amp any day over a gainclone. Others like to listen to Icepower and Hypex, there is room for many approaches.
Ciao T
I am quite familiar with it. While the technological approach differs minimally, in the end we are dealing with a looped DS Modulator.
The same can be said of 1930's Tube Amplifiers.
I have heard them on many occasions. I so far decline to use them. A gainclone, assuming the limited power available from one suffices is a better choice to my ears (and is pure Class B BTW).
But to each his own. I'll take a nice tube amp any day over a gainclone. Others like to listen to Icepower and Hypex, there is room for many approaches.
Ciao T
I'm shocked to see such an obvious failure to apprehend the nature of class D from you ThorstenL.
It just goes to show how a person will make those assumptions that will bolster their prejudices, and how prejudice-driven you are, despite your efforts to present yourself as an evidence-based engineer.
Transitions in a class D amplifier can occur at any point in continuous time, it is not a discrete-time device.
w
The thing that makes me laugh about the Levinson class D amp the most is the sentence that reads.
Excuse me? 135 pounds? That weighs almost as much as I do, that's not playing on any of class Ds strengths. A 500 watt @ 8 ohm, 1000 watt into 4 ohms that weighs 10lbs would be playing on class Ds strengths.
Excuse me? 135 pounds? That weighs almost as much as I do, that's not playing on any of class Ds strengths. A 500 watt @ 8 ohm, 1000 watt into 4 ohms that weighs 10lbs would be playing on class Ds strengths.
How about 25lbs and 4800w into 4Ω?
I think that would be playing to the strengths of class d.
Regardless of quantity.. horrible is still horrible....
I think that would be playing to the strengths of class d.
Regardless of quantity.. horrible is still horrible....
Horrible is not the appropriate word. I have listened to a number of Class D
amps, I own a few, and I've designed a couple. None of them caused blood
to spurt from my ears.
I think the word is uninvolving - Class D amplifiers don't make me want to
run through my music collection.
😎
Yep... alike food without salt.
stérile thing filled with a lot of emptyness.
ahahahahah!🙂
Carlos
stérile thing filled with a lot of emptyness.
ahahahahah!🙂
Carlos
- Status
- Not open for further replies.