Hi, I need to know if class D amplifier are only recommended for subwoofers. Can they also play middle and high frequencies as well ? Can I connect a 4 ohm 3-way speaker (with a woofer, mid-woofer and tweeter) to the output of a class D amplifier ?
Sorry for my nooby question, I did not find a sticky thread for general questions about class D amplifiers.
Sorry for my nooby question, I did not find a sticky thread for general questions about class D amplifiers.
Well I did not mean that, it's just I did not know if they could sound good at high frequencies or not. But tell me something about my other question. Can I connect several speakers in parallel to the output of a class D amplifier ? Of course, respecting the total impedance value. And I am not talking just about connecting subwoofers, what if I connect a subwoofer and a mid woofer with its tweteer to the output of the class D amp ?
Last edited:
Yes, you can use a class D amplifier for full range. and yes, they are used in many sub-woofers brands too. Would I use one for my main amplifier?...NO!
Some do, but many are poor. (..and it is not just about high frequencies)
Simply read through the reviews - and find your taste.
Besides reading you should watch out for possibilities to listen to different designs. The differences are as large as they have been in the early transistorized class AB amps during the 1960s... or more optimistically like in the 70s...
In the mean time the ongoing evolution of class D amps is driven by the goals to achieve low cost and high power density.
Sonic improvements are of lowest interest for generation smart phone.
Consequently the situation of very different sonic quality is likely to last for decades.
The better (Hypex or IRS2092 designs) Class D amps compete directly with the very best Class A amps. The lack of crossover (zero crossing) distortion and (typically) very low distortion overall means that they perform better than most (all but the best and most expensive) Class AB amps. Class D amps are rapidly taking over in the pro world, and not just because of their light weight and high efficiency . . . they sound good too.
That said . . . not all "consumer" Class D amps found in the typical HT receiver rise to that standard (of course the Class AB amps in those receivers are not necessarily "the best" either). And Class D can be (but isn't necessarily) somewhat more "load sensitive" and not respond well to speakers with poorly designed crossovers, although worth noting is that many Class A and Class AB amps have problems with low or variable impedance loads as well.
Of course there will always be some bottle heads out there to tell you that all "solid state" amps suck and the only way to get "good sound" is with some particular brand of EL84s or a single ended 2A3. Don't believe it . . .
That said . . . not all "consumer" Class D amps found in the typical HT receiver rise to that standard (of course the Class AB amps in those receivers are not necessarily "the best" either). And Class D can be (but isn't necessarily) somewhat more "load sensitive" and not respond well to speakers with poorly designed crossovers, although worth noting is that many Class A and Class AB amps have problems with low or variable impedance loads as well.
Of course there will always be some bottle heads out there to tell you that all "solid state" amps suck and the only way to get "good sound" is with some particular brand of EL84s or a single ended 2A3. Don't believe it . . .
Class D has multiple distortion mechanisms, which do not exist in class AB.
And one of these is the dead time distortion, which causes distortion at the current levels were the output current starts to exceed the filter ripple current.
The related distortions are as ugly as crossover distortion.
-Then the distortion caused by the residuals, which are fed back to the input.
- ...further less dominant effects...
IMHO there is no simple reason why classD naturally would sound better.
In fact I am a fan of classD and fell in love with a specific one, which simply kicks out all my older class AB amps with a smile.
One problem with class d is it can be a bit of a black art.
1/ Good pcb layout with powerplanes is vital due to high frequencies.
2/ Correct decoupling to mosfets, power supply and class d IC is vital.
3/ Output filter design needs to be spot on with correct inductor core type.
I fell down all these holes on my first attempt and it just kept on resetting on any level of power. I took 3 pcb revisions to get a amplifier that didnt reset on high power levels,
I can see where poor examples of class d amplifiers come from by my own problems.
I use the irs2092 and had a major problem with clicks, siren noises and thumps on power down when I used 20,000uF per supply rail. I had to contact IR to get help on this.
The solution was to add a small PIC with A2D and monitor VCC and if it was too low hold the 2092 in reset.
1/ Good pcb layout with powerplanes is vital due to high frequencies.
2/ Correct decoupling to mosfets, power supply and class d IC is vital.
3/ Output filter design needs to be spot on with correct inductor core type.
I fell down all these holes on my first attempt and it just kept on resetting on any level of power. I took 3 pcb revisions to get a amplifier that didnt reset on high power levels,
I can see where poor examples of class d amplifiers come from by my own problems.
I use the irs2092 and had a major problem with clicks, siren noises and thumps on power down when I used 20,000uF per supply rail. I had to contact IR to get help on this.
The solution was to add a small PIC with A2D and monitor VCC and if it was too low hold the 2092 in reset.
A class D amp has a minimum load impedance specification, just like any other amplifier. Some are capable of driving very low impedances; for example, Hypex NCOREs are specified to drive 1 ohm, although the lowest recommended load is 2 ohms.
No. It could have been in the beggining of class D amplification, when they did, in fact, have trouble reaching both ends of the spectrum. But the technology progressed.
Today, some of the best speakers available (and the best one I've heard so far, Cabasse L'Océan) are powered by ICEPower class D modules. Excellent sound from Devialet D-Premier (paired with YGA Carmel and Focal Diablo Utopia), and Micromega's amps do a good job as well.
Today, some of the best speakers available (and the best one I've heard so far, Cabasse L'Océan) are powered by ICEPower class D modules. Excellent sound from Devialet D-Premier (paired with YGA Carmel and Focal Diablo Utopia), and Micromega's amps do a good job as well.
This may be an issue with some Class D topologies, but I do not see it with either Hypex or IRS2092 designs. With them managing "dead time" is a thermal issue, but is it producing associated distortion in the audio pass band? Distortion of what kind? At what level? If you have references discussing that I'd like to see them . . .
Since the amplifiers mentioned are completely feedback driven their distortion levels are almost completely determined by comparator sensitivity and noise.
That's my understanding too, but I'm always open to learn something new . . .
Dead time distortion is nothing mystic, but a very fundamental effect.
Unpleasantly it cannot be cured by using high switching frequency.
...I don't want to spoil the thread with to much detailed tech discussion, so I am attaching a LT spice sim file and two PDFs without writing here about all detailed phases per switching cycle.
Everybody can use the attached simulation file and step through the details in order to gain a better understanding.
Similar like in classAB amps the transfer function of the output stage is not fully linear. Class D is suffering from a slightly reduced output at small signal levels, similar like class AB amps with much to low idle current (==> ref. also to class B amps and current dumping approaches). In Feb 1971 Dr. Bernard M. Oliver has settled a very nice article on this topic in the Hewlett-Packard Journal.
Note - in class AB the characteristic also can be opposite, depending on chosen component values and idle current setting.
In order to understand the effect in class D amps, you have to look into the details of the switching transitions. Watch the nano seconds at different current output levels. Key points is when and how fast and with which shape the center node of the half bridge changes from one rail to the other.
Please also note that the simulation is simplified to a basic school book level in order show the dead time distortion.
The dead time is chosen long (almost 100ns) in order to magnify the effect.
It keeps out all the other class D difficulties. It's not a design guide, but highlighting one specific topic only.
Unpleasantly it cannot be cured by using high switching frequency.
...I don't want to spoil the thread with to much detailed tech discussion, so I am attaching a LT spice sim file and two PDFs without writing here about all detailed phases per switching cycle.
Everybody can use the attached simulation file and step through the details in order to gain a better understanding.
Similar like in classAB amps the transfer function of the output stage is not fully linear. Class D is suffering from a slightly reduced output at small signal levels, similar like class AB amps with much to low idle current (==> ref. also to class B amps and current dumping approaches). In Feb 1971 Dr. Bernard M. Oliver has settled a very nice article on this topic in the Hewlett-Packard Journal.
Note - in class AB the characteristic also can be opposite, depending on chosen component values and idle current setting.
In order to understand the effect in class D amps, you have to look into the details of the switching transitions. Watch the nano seconds at different current output levels. Key points is when and how fast and with which shape the center node of the half bridge changes from one rail to the other.
Please also note that the simulation is simplified to a basic school book level in order show the dead time distortion.
The dead time is chosen long (almost 100ns) in order to magnify the effect.
It keeps out all the other class D difficulties. It's not a design guide, but highlighting one specific topic only.
Of course it's not . . . it's two switches. But output stage linearity hardly matters, because what is linear is the relationship between the output and the input, which has to be linear because the whole amplifier is driven by the difference between the two. That removes "output stage transfer function" nonlinearity from the output signal. It's not just a little "negative feedback" to clean up some output stage nonlinearity . . . feedback runs the whole show.
Feb 1971 was a VERY long time ago. How about some actual measurements of this alleged "distortion" taken from actual current Hypex or IRS2092 based amps?
I'll admit not getting what he's talking about. If you open the feedback loop (on a 2092 based amp) it will try to slam to the rails . . . the whole thing is driven by the comparator, not by the input signal.
1971 was the article about the crossover distortion in class B amps.
ClassD:
If one refuses to look into the basics of switching transitions one cannot understand the mechanism of dead time distortion.
That's why I attached the file.
The commutation events of half bridges are the basics from every 4 quadrant switch mode power supply or motor control.
Feedback is helpfull, but the achievable amount of feedback at high frequencies is limited. First one should look the timing details of the switches.
Hypex has a good dead time adjustment and heavy feedback.
IMHO Hypex is one of the companies who really know what they are doing.
This applies for their technical solutions and marketing strategy.
ClassD:
If one refuses to look into the basics of switching transitions one cannot understand the mechanism of dead time distortion.
That's why I attached the file.
The commutation events of half bridges are the basics from every 4 quadrant switch mode power supply or motor control.
Feedback is helpfull, but the achievable amount of feedback at high frequencies is limited. First one should look the timing details of the switches.
Hypex has a good dead time adjustment and heavy feedback.
IMHO Hypex is one of the companies who really know what they are doing.
This applies for their technical solutions and marketing strategy.
"Feedback" in an IRS2092 is hardly "limited", since it is the difference signal that drives the amp. And it happens at ~400kHz. Lots of "bad things" can happen at 400kHz (that's why trace length and component placement is so critical in Class D amps) but those bad things tend to toast the output stage, not to fold back into the audio pass band.
The IRS2092 also provides dead time adjustment, to allow for different speed output switches and prevent failure thereof. It might even have made a difference in audio band distortion back when the switching speed of available devices was itself barely higher than the audio signal itself . . . but that was a long time ago.
Will you please provide some real world (not "simulations") quantified measurements of this audio output "distortion" taken with either Hypex or IRS2092 based amps? Perhaps if I knew when it's supposed to be happening or what it's supposed to sound like I'd be able to identify it in my Class D amps. As it is I hear no difference between them and good Class A or the less common properly biased (LM3886, for example) Class AB.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.