George, I looked at the link you provided. Yes they use filters for the purpose of measurements to eliminate residue of switching frequency that is otherwise completely benign. This small residue (about 1%) has no effect on the sound. Speaker membrane will react (move) to its average value only.
I don’t buy the "gas guzzler" A/B vs. "green" class D arguments for the latter. Kilowatt hours are cheap and relatively clean until you pile up a WHOLE lot of ’em. With typical audiophile usage patterns, and the fact that we’re a VERY small segment of the population, there’s no significant ecological impact here.
According to this pouring used car engine oil to kitchen sink is OK (does not have environmental impact) since very few people do it.
And as far as ecological impact, class D probably makes things worse overall, due to the fact that audiophiles ship ’em over, decide they don’t like the sound, and then ship ’em right off again
That's funny. |
Non-linear & lots of distortion that needs to be cleaned up with an output filter.
I don't understand what you mean by that. Class D is extremely linear. Class D modulator, converting voltage to duty cycle, resembles Delta-Sigma D/A converters (principle of operation is the same) that are extremely linear. Output filter, Zobel network, is only used to obtain average value and reduce amount of noise on the speaker wires (that is inaudible). Like with any amp - don't guess and just listen. I like sound of my small class D Rowland amp. I like imediacy, transients, liquid midrange, low noise floor and composure under loud passages (regulated power supply). The only problem I had initially was lower midrange. It was a little thin sounding. I replaced speaker cables with Acoustic Zen Satoris and lower midrange came back. Cello sounds like cello and I can hear "chestiness" in male voices. Perhaps the question is not whether class D is worse or better than class A (Jeff Rowland makes some wonderfull class D amps), but rather if it is better for you, for the money you can spend. Switching technology is very promising - just look at SACD - pretty much the same as class D. I agree with George that switching frequency should be increased, since 50-60kHz bandwidth limitation introduces around 20deg phase shift at 20kHz (causing wrong harmonics summing), but I'm not even sure I can hear such nuances. There are many highly praised amps that have similar bandwidth, I'm sure. |
Mapman, high frequencies are different for sure, but not for the worse. Cymbals, for instance, sound more "brassy" and less "splashy". To me it sounds more natural, but it might create false impression of limited extension. |
Again, switching noise is inaudible, unless anybody can hear 500kHz. Even bats won't hear it. I'm not sure why is it so difficult to understand. I addition tweeter's membrane won't move at all. Personally, I have problem hearing 20kHz. As for the bandwidth - in my amplifier it is set to 60kHz -3dB. Different sound, more tube like, of class D amp can be explained by low TIM distortion, often responsible for the sharp splashy sound (added higher order odd harmonics). Class D is in fact so clean sounding that sibilants, for the first time, still strong are very clean and wonderful sounding on every record.
Somehow people cannot comprehend that class D amp is purely analog and switching itself is inaudible. Switching is present all the time - even without input signal. Place ear against the tweeter and you will hear very low typical amp's hiss - likely much less than with any class AB amp.
From the sitting position nobody will be even able to detect that amp is ON. |
George , I did put my ear to the tweeter and I can hear almost complete silence - much better than class AB amp I had before. I'm not surprised since tweeter membrane has no chance to move at 500kHz.
Modern SMPS are much quieter than linear power supplies that, in reality, are very primitive noisy switchers operating at 120Hz. Modern SMPS switches at zero voltage/zero current while much ripple is much easier to filter out than 120Hz, that requires huge amount of capacitors (that produce unwanted inductance that is in series with the speakers). In spite of all capacitors linear power supply in power amps is unregulated. Because of that Jeff Rowland does not use linear power supplies at all. He uses SMPS even in preamps to lower noise floor. Benchmark replaced linear power supply in their latest DAC with SMPS lowering noise floor by 10dB.
Mapman is right - you're not sold on class d yet. I enjoy my class D amp immensely. Efficiency is not that important to me (but it should), but is always welcomed as added benefit.
|
|
|
Mapman +1. People are tossing terms like "nonlinear" or "lots of distortion" without even understanding the principle of operation. I can understand when people don't like sound of particular class D amp with their speakers, but tossing pseudo-technical terms is plain silly while starting thread using such terms is ill minded. |
Sorry, Randy-11 but this article is garbage. The critical fact to bear in mind is that all Class D amplifiers have outputs that are comprised of discrete power increments, a notion that can more easily be thought of as resolution steps.
Class D amplifiers have no "steps" at all. They are analog with unlimited resolution. Author cannot comprehend that in class D amps quantity of interest - voltage is replaced, in linear modulator, by duty cycle . Amplifier jitter rejection can also be critical
There is no such thing. Analog amplifiers with voltage input have no jitter rejection. Author appears to be uneducated thus unfit to write on the subject. |
Nonsense. Not only that I don't know what they mean by "Pumping back the energy" in half bridge configuration (that I use all the time) but at least they don't mention limited resolution. Such statements: However,
the differential output structure of the
bridge topology inherently can cancel even the
order of harmonic distortion components and
DC offsets, as in Class AB amplifiers. A fullbridge
topology allows of the use of a better
PWM modulation scheme, such as the three
level PWM which essentially has fewer errors
due to quantization. are gibberish. Just in case your'e interested Icepower is Full Bridge while Hypex is Half bridge. Which one is better - let your ears to be the judge. Full bridge places half of the supply voltage on the speaker wires all the time. Look at incredibly low THD and IMD specifications (and measured data) of class D amps - not possible with non-linear amplifiers. In all this discussion about errors (power supply voltage or dead time) author forgets about negative feedback (dual in Icepower amps). Without this feedback class AB amps would sound like shit. Their conclusion: Highly efficient Class D amplifiers now provide
similar performances to conventional Class AB
amplifier |
I will state just the opposite. Violin sounds like violin without scratchy glazed harmonics of class AB amp. Same goes for the cymbals that sound brassy and meaty instead of bright and splashy. Cello sounds deep and resonant. Transients are incredible while amps keep composure under heavy current demands (orchestra forte), most likely due to regulated power supply. The other thing I noticed is low level performance. When I listen soft at night - highs and lows are still there. Imaging is even better, but it might be due to reduced reflections.
Somebody described sound of Hypex as a sound of very good class AB amp while Icepower was described as more of a tube amp sound (half bridge vs full bridge output). Icepower is very good in spite of some negative opinions here. Now we have more choices, Pascal being one of them.
Speed of Mosfets improves every single year, but now new switching devices are coming - like Enhanced Mode Gallium Nitride Fets (eGaN Fet) that switches many times faster without overshoot or ringing due to ultra low capacitance and inductance. The best days of class D are ahead of us, but I enjoy it immensely already. |
Just because you heard a class D amp sound better than a tube amp in a particular situation does not mean that all class D amps sound better than all tube amps, or even that the particular class D amp sounds better than all tube amps. It just means that in that situation you heard something better that you liked between the two involved.
Who said that? That statement is completely erroneous Who said? you know exactly who, since you copied. This statement, made by Atmasphere, is absolutely true stating that you can like any class of an amp more than another. Nothing erroneous about that. |
If bought down low to erase most of the switching noise, this intrudes into the upper frequencies of the audio band. Which makes it very smooth but robs most of the harmonic structure of the mids and highs, this "could" be compared to tube or class A s/s smoothness. My class D amp has 60kHz -3dB bandwidth. Georgelofi - you must be a bat? |
Mapman +1.
Georgelofi - if you don’t like the sound of particular amp or even class of them just say it, but making up pseudo-scientific reasons is just plain silly. |
Rowland model 102. There are many class A or AB amps that have similar bandwidth but nobody talks about phase shifts.
|
Georgelofi - you are talking about harmonic distortion below 0.1%, that are inaudible (unless you're a bat, of course). I've noticed that you also tried to apply psuedo-science to sigma-delta converters. I am afraid one day you'll find that SACD is the same.
|
The output filter vs switching noises frequency is the bottleneck for Class-D HF distortion and low impedance drivability. It is ill minded and made up. There are many amps that have similar bandwidth and great reviews. My tiny amp is listed as 3 ohms minimum and it drives 3.6 ohm min. Hyperion HPS-938 speakers. Larger modules, like one used in mapman's amp, can drive 2 ohm (this module can deliver 40 amperes for 0.5 sec and 36 amperes for 1s - not many amps can even do that). One of our audiogon friends uses class D amp (H20) with Apogee Scintilla in 1 ohm mode with great results. Please stop making pseudo-scientific reasons - it is not funny anymore. |
Before you offer your "expertise" get things right - mhz is millihertz, proper unit is MHz (megahertz). Keep posting - you might learn something one day. |
It is far from mud slinging, buy it appears that you make strong technical statements on the subjects that you are completely unfamiliar with. I pointed out units just to show that you don't have any electrical engineering education. This thread is not too bad, but your crusade against Delta-Sigma converters is plain insane. Cheers. |
It happens that this person (muralman1) knows everything there is to know about Scintillas and he adores them with class D amp. Nice try :(
|
Phase shift and wrong summing of harmonics argument might look plausible (I claimed it myself) if not for the fact that my speakers already limit bandwidth to 22kHz creating larger phase shifts in comparison to 65kHz amp's bandwidth. This 65kHz amp's bandwidth is responsible for about 20deg phase delay at 20kHz (that I cannot hear) and about 10deg delay at 10kHz (that I can hear). One can argue that this phase delay might be beneficial since many speakers have positive phase shift at these frequencies. I don't have data for my speakers, but I picked one of the speakers reviewed by Stereophile. As you can see on the phase diagram (dotted line) speaker actually accelerates phase at the highest frequencies. Phase delay caused by the bandwidth limit of my amp would improve summing of harmonics (Al, am I missing something?). http://www.stereophile.com/content/monitor-audio-platinum-pl300-ii-loudspeaker-measurements#Oze7AYHq...There are many class A or AB amps that have bandwidth much lower than 200kHz (including Krells that have -3dB at about 100kHz) but nobody makes claims about bad sounding highs. It is a little of negative placebo effect - if you really believe that class D cannot sound good it will not. As for the hefty H2O power supply - it is unregulated and as such requires huge bank of capacitors to keep voltage steady and to filter out 120Hz ripple. Jeff Rowland uses only very quiet SMPS, line and load regulated that operate at 1MHz switching frequency that is very easy to filter out. In fact, he uses SMPS in preamps (Capri), where efficiency is not important, to lower the noise. SMPS got bad rap from crude computer applications. Either way works but traditional "linear" supply generates a lot of switching noise. Also low inductance capacitors (like slit foil) are very expensive. Why only a few companies use SMPS? Perhaps because it is not easy to design good switcher and also because of market demand. Many people believe that it has to be very heavy to work right. Tiny ferrite transformers can carry at high frequencies as much power as huge transformers operating at 60Hz - not to mention wide supply voltage range and DC operation. SMPS in mapman's amp can deliver 36 amperes for a full second! |
|
Here is part of Lamm Industries ML3 Signature amplifier by Michael Fremer in 9/2013 Stereophile. The ML3s produced the most glorious, palpable, airy, detailed midrange I’ve ever heard from reproduced music. That part is easy. They pushed that performance envelope without going all greasy and congealed over time, as some tube amps do after the initial appeal of warmth wears off. The ML3’s standout features were its natural re-creation of instrumental attacks, generous sustain, and lifelike decay—all as close to live as I’ve heard, if nowhere near the real thing. It was quiet, fast, detailed, dynamic without reservation, transparent, airy, and extended on top. No sharp edges unless the recording had them, and no boredom-inducing global softening. The pair of them produced an enormous sense of space when that was appropriate, and, within that space, images of exceptional delicacy, three-dimensionality, and body.
It should be good for $140k/pair. This amplifier has measured -3dB bandwidth at about 50kHz - which is 15kHz less than my class D amp, that supposed to suffer from wrong harmonics summing. Please notice words "airy, and extended on top". Imaging also did not suffer. |
That's general idea, but simple it is not. None of current class D amp modulators uses triangle waves. If anything it is more sinewave and closely resembles modulator use in Delta-Sigma converters, that produce similar output. SACD is also a product of such Delta-Sigma modulator and it is basically class D output signal (PWM). In addition to modulator alone there are different configuration of output stages, output filters, multiple feedbacks etc. Let just say that it is so "simple" that it took whole PhD work of Karsten Nielsen (Icepower) at University of Denmark to describe theory behind it - much less actual implementation. Once you take FPGA approach to class D you will find that it is very, very complex. |
Speakers shift phase much more (in opposite direction) at higher frequencies. Imaging should be OK as long as phase in both channels match (no reason not to). As for the switching noise - I cannot hear 500kHz, but I'm pretty sure some people here will claim they can. |
Nobody can hear 500khz, it the filter that has to deal with it and what’s left over and the byproducts of it that is the problem.
What is left over is 500kHz. Byproducts are possible (modulation), assuming nonlinearity of motion, if tweeter can move membrane at 500kHz - no chance. Filter or frequency have very little to do with quality of the sound, IMHO. It is related to quality of the modulator, slew rate of output devices (accuracy of the duty cycle), ringing caused by inductance and capacitance of output devices etc. Actually, increasing switching frequency can make it worse because it will reduce accuracy of the duty cycle - calling for deeper NFB. As I showed before (using 50kHz Lamm amp as an example) phase delay in audio band has nothing to do with extension and, if anything, might be beneficial since most of the speakers accelerate phase (much more) at these frequencies. Also, increasing switching frequency will reduce amount of residue on the speaker cable but at the same time it will make it radiate much better. |
Well, most of people on this forum (including you) think that SMPS is noisy and linear power supply is better. This believe comes from bad rap of noisy crude computer SMPS applications. They don't even realize that "Linear" power supply is a primitive unregulated switcher. I addition, many believe that when something is smaller and more efficient something else has to give (that's human nature). They apply this believe to both SMPS and class D amplifiers (that are basically controlled SMPS). In view of that amplifier manufacturers would be stupid not to keep using "Linear" power supplies. Only few true pioneers like Jeff Rowland had courage to develop state of the art SMPS supplies that are extremely quiet (noise very easy to filter out at 1MHz operation). Even 50kHz SMPS executed well are far better than "Linear" supplies IMHO. |
This is a huge step in the right direction getting rid of another point of noise (the smp), with the added lift of future higher switching frequency/filters and the benefits it brings, would spell the end of all linear amps. And Srajan's tubes below.
Practically all supplies are switchers. So called linear are very primitive -switching at 120Hz when voltage is the highest, creating a lot of high frequency noise. Current is drawn in narrow spikes of high amplitude. SMPS on the other hand, when done right, is far better in many respects. It is quiet, switching at zero voltage/zero current. It operates at high frequency that is easy to filter out (much easier than 120Hz ripple) and is line and load regulated (vs. unregulated linear supply). It has fast response to load demands - vs. slower response caused by inductance of huge electrolytic caps. It tolerates DC on the mains (It even operates from DC). Rowland switched completely to SMPS using them for class AB amps as well as for preamps (to lower the noise). Benchmark improved S/N ratio by 10dB by replacing linear supply in their DAC with SMPS. New class AB amp from Benchmark ABH2 also utilizes SMPS resulting in 132dB dynamic range. Why then, designers still use linear supplies with huge transformers and capacitors? For two reasons: First, it is very difficult to design good switcher. Second, they follow market demand - many people believe that it has to be "linear" and heavy. |
The main reason manufacturers of high end audio (like Rowland) use SMPS is low noise and line/load regulation. You can find it in FAQs on his (or Benchmark) website. He uses SMPS in preamps where size or efficiency play no role.
Yes, top manufacturers still use linear power supplies, but main reason is demand from people who, mistakenly, believe that if it has to be heavy to be good. |
Speaker cables would be a very poor antenna. Since wavelength of 500kHz switching frequency is 600m a typical 2m cables would be 1/300 wavelength antenna. Usually antenna becomes rapidly ineffective below 1/10 of the wavelength. |
Maybe we will start seeing Class D preamps then...
Perhaps. Delta-Sigma D/A, A/D converters, SACD and DSD are already class D. |
There is about 1% of switching voltage noise on the speaker cable since filter is not perfect but, as I said before, 1/300 wavelength antenna won't radiate. In addition, electricity in speaker cable flows in both directions canceling the most of RF (if any). When you place radio directly on the wire you might get some capacitive coupling, but I doubt it. Many speaker cables are twisted making it completely inefective (better than shield) for electromagnetic radiation or capacitive coupling (both ways - from cable or to cable). |
Talking about Spread-Spectrum during cold war could put you in a prison (top secret). Today all CDMA cellphones (Verizon, Sprint etc.) use spread spectrum. It is interesting application for class D, especially when receiver reacts only to average value (no keying needed). |
When electricity flows in two wires in opposite direction it tends to cancel radiated EMF or capacitively coupled noise. Same goes for receiving - the closer two speaker wires are - the less susceptible they are to pick up anything. Twisted pair exposes both wires symmetrically to external fields making them almost immune to coupled or radiated noise. That's why in any circuit design signal and return should always go closely together to avoid any loops. The point of my post was that cable (two wires where electricity flows in opposite direction) is a very poor antenna for radiating and receiving. Twisted pair is practically no antenna at all. |
Perhaps this explanation will be easier to understand: If you make loop of a wire conducting electric AC current the radiated EMF will be proportional to loop area. If you reduce area to almost zero (like wires together in a cable) the radiated EMF will be almost zero. |
Stereophile test shows about 0.5Vp-p - that is about 1% since switched voltage is in order of 50-80V. It comes from the fact that output filter is 2-pole making it -40dB/decade while switching frequency is about decade above filters cutoff frequency suppressing noise by 40dB. -40dB is 1%
As I posted before, speaker cables have no chance to radiate for many reasons. Tweeter won't play 500kHz, not only because of membrane inertia but also because of its impedance rising extremely high.
|
Switched voltage 50-80V represent the highest output voltage. They show switching noise riding on a very small signal. This signal of 1V peak represents 63mW of the root frequnecy + harmonics on the 8 ohm speaker.
|
You might damage membrane from overheating, but most likely not from 500kHz switching frequency. As for switching frequency level following signal level - your understanding of class D output is poor. Amount of switching noise is constant independent of the signal level and in your Stereophile example equivalent to power of 4mW (of inaudible frequency) on 8 ohm speaker.
This switching noise is only dependent on the switching frequency and the filters cutoff frequency. Amplitude of the music signal has nothing to do with it!!! If anything, switching noise at extreme duty cycles will be lower. Signals of 10% or 90% duty cycle represent lower amplitude (in comparison to no signal represented by 50% duty cycle) of the root frequency (500kHz) and a little bit higher harmonics. After filter amplitude of 500kHz will be lower (and not higher) at high music levels.
|
I'm not sure what NuForce are doing. It is possible that they adjust supply voltage to signal level, but that would only mean much lower noise (than for instance Icepower) at low signal levels. Noise level at max loudness would be the same as for the amp that doesn't do that (assuming same switching frequency and the filter) - equivalent to about 4mW on 8 ohm speaker. Icepower doesn't adjust the rails and I'm pretty sure Hypex uses constant supply voltage. With constant supply voltage amount of switching noise at 90% or 10% duty cycle would be most likely 1/2 or even 1/3 of the one at 50% duty. |
AFAIK, class C is a poor choice for audio (better for FM transmitter).
|
In learning about Class D, it appears that after the linear signal is destroyed and turned into a saw tooth wave form, the output filter? tries to put the signal back together...doing an incredible job considering the complexity of this task...but ultimately is unable to restore micro details to recreate the original signal...which people such as myself notice very quickly, while others either don’t notice or enjoy the new coloration.
It is silly, to avoid using stronger word. You originated this thread pretending you want to learn about class D, while knowing you hate it. I will remember your login and will be careful to answer your posts next time. |
No, learning about class D is not your goal. You placed many negative posts on class D with no single positive one. That is trolling. |
I am not sure where obsession with the switching frequency comes from. It is inaudible while the only other thing that it might cause is filtering network phase shift - that can be beneficial since speaker has much larger positive phase shift at high frequencies. Placing nonsense about signal being "destroyed" (while SACD works exactly the same way) is trolling.
In fact raising switching frequency might be harmful because of the dead time, forced by the controller to avoid shoot thru, that Atmasphere mentioned. Since this time is fixed, error it makes will be larger at the higher frequency. One might not like the sound of particular class of amplification but making pseudo-scientific explanation i plain silly. |
|
Ok, let me clarify then: I cannot hear 500kHz, not only because of my hearing apparatus (at my age I have difficulty hearing 15kHz), but also because tweeter’s output drops like a rock with combined effect of membrane inertia and increasing tweeter’s impedance (easy enough to check tweeter’s frequency response chart). There might be intermodulation products, but only if tweeter’s membrane can move at 500kHz - no chance . In addition output filter, set to about 1/10 of the switching frequency, is at least two pole resulting in >40dB/decade suppression. 500kHz frequency that I cannot hear and tweeter cannot response to is already suppressed by >40dB from the peak loudness. I challenge anybody to detect, form the seating position, when my amplifier is on.
|
I'm not sure what this "class A input stage" is. Input stages are almost always class A (would be stupid not to), but class D amp is pretty much one stage - a modulator + output transistors. Any additional input stage is not advantageous to class D amps. What they call "Analog Cell" is just class A Mosfet input stage very common in many class A or AB amps. It sounds like a lot of hot air (for class D). |
Perhaps in traditional multistage amps. I'm not also sure what the "voicing" means. Would this imply that two different amps with the same Mosfet input stage are voiced exactly the same? As I said - a lot of hot air. Manufacturers of expensive gear have to come up with some exotic names, like "Analog Cell", in order to differentiate them from competition and justify high price. Icepower is self contained amp. It doesn't need additional input stage unless it serves a purpose (Common mode noise rejection, input impedance increase etc), or "voicing" is just a fancy name for added "coloring" of the sound (at the expense of transparency). |
Yes, sometimes there is additional circuitry to Icepower modules. For instance my small Rowland has additional instrumentation amp (THAT1200), that improves common mode noise rejection as well as increases standard 10k input impedance to 40k. Box is pretty indeed, but since it is solid billet aluminum chassis - board inside is acoustic resonance free. There is a purpose to that - but when manufacturer puts Mosfet stage in a plastic blue box, calls it fancy name "Analog Cell", describes it as class A with no feedback and inserts picture of the board with this blue module in their literature - then it makes me very suspicious. Not only that using "class A" is to pray on non-technical people, but also second statement is not true. There is practically no stage without feedback. Even resistor in the emitter (or source) of transistor is a form of the negative feedback. First stage might be outside of the global negative feedback, but calling it "no feedback" is again praying on the common believe that "no feedback" is a good thing. Because of all that I'm very suspicious about additional Mosfet stage used for the "voicing". We also have to assume, that B&O engineers, that are fanatic about sound, and Jeff Rowland neglected to properly "voice" it. Would people buy this amp, after finding in the sales literature, that the company just packaged inside standard inexpensive Icepower modules? Most likely not and that's whole reason for the "puffing". http://www.psaudio.com/stellar-m700-mono-amplifier/ |
+1 noble100, I was an enthusiast of class D amplification from the start. I replaced my (medium quality) class AB amp (Cambridge A3i) with one of the first class D amps (Rowland 102) based on the one of the cheapest Icepower modules (200ASC) and was very impressed. Recently I replaced it with class AB Benchmark AHB2 rated class A in Stereophile. It is a better (wonderful) amp overall (cleaner, better extensions etc.), but the difference was small, taking into consideration 3x higher cost (in comparison to Rowland that was well made and expensive). Class D technology is only going to improve in time while everything else is pretty much done. There are still some new designs, like my Benchmark AHB2 amps, that dissipate less and have non-recursive NFB, but class D is the future IMHO. |
