Very interesting posts!
Most, if not all, class D amps put out a lot of noise starting around 15 to 20 KHz and climbs to surprisingly high levels in the ultrasonic and up into the RF range.
This noise can be filtered out if the amp is designed into a dedicated speaker. But in the case of a free standing amp, such as the unit I have, there is little that can be done about it since the designer has no idea of what speaker will be connected to it. The impedance curve of the speaker has to be taken into account when the filter is designed and therein lies the problem.
Eldartford does have a good idea to see if the Class D amp is messing with the upper ranges in the main speakers. I will have to try this now that I am curious about it. But as Sidssp pointed out, my question was more oriented toward how the Class D amp sounded in the sub bass, where it is driving the speaker.
http://datasheets.maxim-ic.com/en/ds/MAX9736-MAX9736B.pdf
Take a look at this datasheet for a Class D amp. Although this is a rather inexpensive Class D chip, you can get an idea of the noise output in the ultrasonic and RF range. Go to page 8 and look at the middle three graphs. Note how much noise is in the inband and outband graphs. This will indeed affect the sound. Just understand that the audiophile gear we buy would have better noise characteristics than this particular chip. Also, as I mentioned above, dedicated amp/speaker combos would have a filter to eliminate this. But this datasheet does give you an idea of the noise from a Class D.
Anyway, I starting to sidetrack. What I noticed about my system here is that the bass is not quite as lively sounding; rather it strikes me as a bit on the dead side. If I had to guess, the somewhat dead sound is due to the noise output.
Before I installed the 450W Marantz unit, I had four Amber 70's driving the four subs and it sounded much more musical. The problem was the 70W Amber ran out of voltage on the louder passages and clipped. Hence my purchase of the Class D - actually, I hadn't planned on buying an amp at the time but the price was too good to pass up! With 450W into 8 Ohms it doesn't clip now, that is for sure!
However, it gets worse.
Both of my tuners, the Amber/Rotel digital and the older Sansui TU-777 are both unusable when the Class D amp is on. There is so much static and interference, the system is simply not listenable. I have tried a lot so far to fix it but nothing has had much of an effect. Here is a brief list: separate AC line, huge ferrites on AC line, rca inputs, and speaker outputs, moving all the cables from the amp well away from everything else, etc. So far, nothing has helped. I did the same thing to the FM tuners, too.
My next idea is a large attic mounted directional antenna with a booster on it. I am hoping the larger signal working along with the internal squelch circuit, might block the noise. After that, my next trick is installing a large ground rod and running #6 copper ground wire to the chassis to see if that helps. I might try grounding the FM chassis, too.
I suppose I could just go out and buy four Mark Levinson, Krell, or Parasound monoblocks but let's keep it real and not dwell in fantasy land.
Anyone have any other suggestions? |
Kijanki - switching noise is typically defined as clock noise in this digital world we live in. Although it is technically correct to say "switching noise" when a diode turns on and off, that noise is insignificant to clock noise on the output of a Class D amp. It is incorrect to say a Class AB linear output stage produce switching noise when moving from positive to negative or back. If it produced switching noise, it wouldn't be linear by definition.
I will have to check to see what Class D noise curves I have left over from when I worked at Maxim. I know I have some white papers showing some noise at 15 KHz.
It really is amazing how much noise there is in the output of a Class D amp. Unless you have a filter on the output, you will have noise, there is no question of it. The only question is at what frequency ranges the noise will peak.
The filters in question here are hardly zorbel networks, although they would help a bit. The filters i have seen in use are usually 3 or 4 poles and are rather aggressive, high Q passive networks.
Depending on the amp, you may or may not have audible noise at 15K Hz. As I mentioned above, comparing the inexpensive Maxim chip to an audiophile piece of gear such as your Rowland is just not comparing apples to apples. Frankly, if your Rowland produced audible switching noise at 15KHz, I would send it in for repair. Just understand that noise at 60 db down will impact the sound, you just won't hear it as noise. My personal belief, based on designing a lot of audio amplifiers, is anything above 140 dBV inside a linear amplifier with some loop feedback will impact the sound in a negative way to some degree.
You don't need a quarter wave antenna to jam the airwaves so that FM broadcast won't get through. I suspect the noise problem I have is more harmonically related than fundamental related. A 1/4 wave antenna at FM frequencies is on the order of a few feet, depending on the propagation velocity of the wire.
Also, my Class D amp is only 10 feet from the tuner and maybe 3 feet from the antenna. The FM broadcast antenna is miles from here. Granted, the broadcast antenna is in kilowatts and the amp is not, but the energy from both decreases as a square of the distance. The amp may not produce as much energy as the broadcast antenna, but here in my living room I wouldn't bet any money on which produced more signal in the antenna.
I don't know if you live near me but a few seconds of having this Class D amp powered on with FM will convince you there is a serious radiated or perhaps conducted noise problem here. As I write this, I have the FM radio on, but the subs are not powered on!
What I do find really encouraging about all of this are the number of folks who are really happy with their Class D amps. If everyone here was telling me I have a problem simply because I was dumb enough to buy a Class D amp, I wouldn't bother to try and fix this problem. |
However, when not listening to the FM radio, the amp sounds perfectly wonderful! I had some classical music on last night - piano, horns, and violins - and the sound was truly amazing! |
I've been out of town for the last four days and just got back. I'll read up on all the posts and write a bit more tomorrow. Thanks for all the responses! |
Kijanki: Sometimes I write in the manner I speak rather than write in a more coherent manner. As an Electrical Engineer, I know the difference between Watts and Kilowatts. I apologize for not being more clear. Frankly, I wish the station broadcast in the MegaWatt range, at least the signal would be a lot stronger here. My favorite Jazz station is privately funded - their broadcast is so limited I can only listen to it clearly in Silicon Valley. East of the hills in the East Bay, their signal is very limited. Sigh...
High Frequency noise is "easy" to filter out within a band of frequencies. However, as the bandwidth increases, filters become more difficult to design and thus noise become more difficult to eliminate. When you consider you also have to filter electromagnetic noise as well as electrostatic and conducted, and the fact our noise spectrum now extends into the GHz range, it is no longer a trivial task.
Do understand that any noise spikes created by the diodes in a linear power supply is insignificant in comparison to the noise created by diodes and components in a switching power supply or class D design. This doesn't mean anything with a switching power supply is bad, but it does mean more engineering work has to be done to eliminate the resultant noise that comes along with it. Also note that this diode "switching noise" in a linear amplifier is actually the transformer radiating, not the diode. But, again, this noise is insignificant in comparison to the noise created by a switching power supply of comparable quality.
Class D amps grossly violate the KISS principal, but it is being widely embraced because most of us who have a multichannel 300W+ per channel Class D amps couldn't afford it if it was Class AB and certainly couldn't afford to operate it, much less buy it, if it was pure Class A. This is in addition to the fact that Class D makes higher power, battery operated personal stereos possible.
Given a fixed amount of money to spend, choosing a Class D amp for the midrange and upper frequencies would be my last choice over a good tube or Class AB solid state design. I do believe it is much more difficult to design an equivalent sounding Class D amp to a Class AB design.
Tgrisham: I agree with you on implementation. This Marantz amp I have had a sticker price in the 2000+ dollar range. I don't remember exactly, I would have to look at the receipt if I still have it, I am thinking it was around 2400. I bought it at 60% off due to store closure and the fact it was only four channels and not six made it hard to sell. Clearly, it isn't a cheap design - even though it is Class D it still weighs over 60 pounds. It is possible that I have a bad amplifier but I am more inclined to believe I have an implementation problem of some sort here at the house. A Class D amplifier design engineer I know now works at National Semiconductor, I think I will look him up tomorrow and discuss this noise problem with him.
Bill: I agree, the use of an amplifier as a subwoofer only makes the choice of amps much more easy. For one thing, it has only a very limited range of frequencies to reproduce and any nonlinearites created by producing 80% of the power in music is not fed to the upper octaves where your ear is most sensitive.
In addition, designing a subwoofer only amplifier is easier in that more feedback can be used without the resultant hardness through the midrange. The additional feedback increases the damping factor which usually improves the bass and the limited frequency range of the bass frequencies means upper bandwidth can be sacrificed for the increase in feedback to maintain stability.
Unfortunately, history has shown it is more difficult to sell a standalone, subwoofer only amplifier. Although once installed in the subwoofer cabinet, they seem to be selling quite well! |
First, let me state I "fixed" the radiated noise from the Class D amplifier interfering with the FM radio. I bought a new antenna and positioned it 20 feet away from the Class D amplifier. Now, there is no audible noise in the FM. I also found that the antenna I used had more of an effect on sound quality than the FM tuner!
Using the standard, cheap, 300 Ohm lead dipole that comes with every FM tuner you buy could produce either good sound or poor sound, depending on how you orient and where place it. Switching to a directional, high gain, wide bandwidth FM antenna on a rotor could make the sound quality better or worse than the cheap dipole, depending on where it is placed and its orientation. I think the next step is to buy a better directional antenna and install it in the attic with a rotor.
There is no question that both of my Class D amps radiate like the dickens and the pickup is through the FM front end. Adding massive EMI cable ferrite assemblies on the coax, line cord, and speaker wires did little to solve the noise problem. |
Kijanki: What you wrote is not exactly correct nor complete. The conduction current spikes you mention do exist in the manner you describe, however your analysis is not complete.
First, the conduction spikes caused by diode action you describe are well known and are published in every engineering textbook that discusses converting AC Mains power to unregulated DC voltage. I am pretty certain the first engineer to build a circuit which did this was in the mid or early 1800's, long before DeForrest invented the triode vacuum tube in 1906, I think it was.
Don't think for an instant that switching noise spikes don't exist in a SMPS, they do and are larger in magnitude than in a linear PS. They are, however, quite different in nature in a SMPS than in a linear PS. In addition to this, you also have noise from the clock frequency to deal with. There are SMPS out there with clock frequencies at 1MHz and higher, although I believe most SMPS run in the 100KHz-ish range.
The method to eliminate radiated noise caused by the current spikes in a transformer, whether SMP or linear power, are the same and is not difficult nor impossible to do. It does add cost however, and that is the key.
In addition, the frequency of operation also comes into play: the higher the frequency the smaller the size of transformer for a given power output. However, higher frequency does not mean lower radiated noise. What it does mean is that the transformer is smaller and thus the shielding can be smaller, meaning lower cost. SMPS's are also significantly less expensive than linear power supplies for a given power. They are also much more complex. Note that the stability requirements for a SMPS are the same as for a linear, regulated supply as both involve loop feedback.
The first approach to remove radiated transformer noise, which is probably the most common, is to design in a Faraday shield in the transformer. A transformer using a torrid design is a better and more expensive approach.
Second, a Faraday can could be placed around the transformer and depending on how much money you want to throw at it, you can make it out of iron, copper, iron and copper, or mu-metal.
Third, you can remove the transformer from the chassis and place it in another chassis several feet away from the sensitive electronics. Certainly this has been done numerous times in high end audio equipment. Fourth, you can use a combination of these approaches, not to mention this is not an exhaustive list. For example, I did not mention the quality of the transformer design.
If you want to remove the current spikes caused by the full wave diode bridge circuit in front of a huge capacitive input filter for a large power amplifier, the best way is to eliminate the capacitor based filter which causes the current pulses.
Instead, design a PI based filter where the series inductor is designed to handle the worst case current the amplifier is expected to deliver. In this case, the current spikes which occur become nothing more than 60 Hz sine waves where the magnitude of the current is about what the power supply is delivering.
Don't misunderstand me, this is still an unregulated raw supply. To complete regulation, you still have to add a linear regulator on the output, which is effectively adding another amplifier in series with the signal amplifier. A SMPS does exactly this, as the SMPS is effectively another amplifier whose output is DC.
I am sure you can see that a SMPS power supply is less expensive than the standard regulation circuit just described. Just as there is no question a SMPS is less expensive, there is also no question a SMPS is more noisy and that noise has to be dealt with.
Note one can use all the noise killing techniques described above, including putting the PS in another chassis, and then using a regulated SMPS as well to reduces the cost over a linear supply.
This is indeed a less expensive approach, but it is still more expensive than using a full wave, capacitive based, raw supply which is why this is still the most common approach in linear power amplifiers. Since the amplifier has inherent power supply noise rejection, and the 120 Hz ripple left over in the output voltage is well within the feedback loop bandwidth, I am sure you can see why this is such a popular approach.
There is no question, at least in my mind, that there are better sounding approaches to power supply design in power amplifiers than an unregulated raw supply, it doesn't change the fact it is still a popular approach. |
Kijanki, I forgot to mention that I agree with your statement that Class AB is not expensive to operate, I disagree with your statement that distortion mechanisms is the reason for Class D in audio today.
The reason Class D is in audio today is the lower cost to produce a given power output and efficiency of operation. Let me discuss these separately, since they apply to different markets.
A Class AB, two channel, 200 or 300 Watt power amplifier is probably affordable by most audiophiles unless one considers the more exotic and expensive designs. Certainly, I don't need to mention the more expensive brand names here. However, given the cost of the more mainstream 200W stereo amplifiers, it should be obvious that a 200W Class D amplifier is less expensive for a consumer to buy than the same power in a Class AB amplifier.
Class D has really come into its own due to the proliferation of home theater systems where one needs 6 or more channels of operation, the cost of these systems are much lower using a Class D approach. In addition, the proliferation of portable sound systems demand the efficiency of Class D operation since they have to operate by battery power and conserving battery power is everything in a battery based design. Battery powered personal sound players is obviously a big market as are laptops computers which also have to produce sound as well.
As pointed out elsewhere in this thread by knowledgeable folks, implementation of a technology is more significant than the technology itself. Granted, a Class AB design has some inherent distortion mechanisms eliminated by a Class A design, that doesn't mean a Class A design sounds better than a Class AB design. Whether it does or does not is dependent on the trade offs chosen by the design engineer and the manufacturing techniques and quality.
The same applies to a Class D design. I disagree with your statement that a Class D amplifier has inherently less TIM distortion than Class AB design; whether it does or does not is irrelevant. Frankly it adds new distortion mechanisms that do not exist in a Class AB design. Given the increase in complexity of a Class D design and circuit, implementation is substantially more significant to the quality of the output signal.
This reminds me of CD player manufacturers which still specify wow and flutter as a distortion mechanism. It is obviously designed out in a CD player but it is rare to see a jitter specification in inexpensive CD players. Jitter is a new distortion mechanism which did not exist in a LPs. |
Kijanki, I don't believe any SMPS is quieter than a linear regulated supply regardless of what the diodes are doing in a linear PS. Sure you can design SMPS to switch on and off at zero crossings but you still have to deal with clock noise. If a SMPS is low noise, it is only because some design engineer somewhere worked pretty hard at getting the noise out.
There are cases where linear regulators are used after SMPS regulation simply to eliminate the SM noise. I think Maxim has a patent on this and they do use this approach in many of their sensitive chips. I know, I worked there as an apps engineer for a while.
Also, please note that any noise, at any frequency, injected in the bandwidth of the amplifier has a negative impact on the sound quality. This is especially true at higher frequencies where PSRR of the amplifier drops off.
The reason for this is that any noise injects additional energy into the bandwidth of the amplifier. Sure, you may not hear it as "noise" but any increase in energy within the bandwidth of the amplifier will sound different. This is the basis Nyquist theory and all digital audio: a given bandwidth and given energy level sounds the same. Change either one and the sound changes.
Again, as I have stated at least three times now, just using a SMPS, or a Class D amplifier for that matter, doesn't mean the sound is worse than an Class AB design, it could be substantially better. It all depends on the trade offs made by the design engineer and how it was manufactured.
Guido: I stated above that I believe the sound of my bass is "a bit dead and not overly lively". I based this on listening to it, not from some datasheet or preconceived made up opinions about Class D amplifiers. I did not state that all Class D amplifiers produce bass that is dead or not overly lively. I did state that this is what I hear in my system and it is caused by my Class D amplifier. I was curious to see if anyone else had experienced the same sonic effects which I have.
Let me put it another way, approximately 1000 folks have looked at this thread. If say 30 to 40% of them stated something like "Oh, heck I have the same problem with my Class D" I would be inclined to find an engineering reason for it. This has not happened. Whatever I am experiencing here seems to be more related to my own Class D amplifier.
If I felt that Class D amplifiers were bad from bias, I wouldn't have spent a grand on Class D amplifier in the first place! I could have continued using the two Amber 70's I had. Actually, the only complaint I had about the 70's is the fact they only produce 70 watts. This is not a lot of power and I can clip them if I was not careful on the volume control. Otherwise, they sounded perfectly fine as bass amps. My subs are nominally 6 Ohms and the Amber's have plenty of current to drive 6 Ohms. The problem is they were running out of voltage, not current. A well designed amplifier should run out of voltage when they clip, well designed amps should not run out of current.
These aren't theoretical generalizations I am making up. If any Class D amplifier is going to have less radiated or conducted noise than a Class AB, it is because the design engineer made it happen because Class D designs have inherent noise that Class AB amps don't have. That is a fact of life, just as tubes have filaments, mosfets are voltage based devices, and transistors are current based devices.
I will admit that I am conservative when it comes to making changes in my stereo - This is the reason I have a Class D amp for the subs and I haven't sold my Premier Five CJs on the upper octaves. I could have sold the CJ's and ran the entire system off this monster Class D amplifier. I didn't do it, I wanted to see how it worked on the subs.
Please scroll up and read what I wrote about radiated noise from my Class D amplifier causing FM interference. If it was conducted noise, moving the antenna would not have any effect. But the fact that I moved the antenna and eliminated the interference proves the noise is radiated. Also, simply turning the Class D amplifier off eliminated the noise also proves my Class D amplifier is the source of the noise. If anyone still believes this noise is not from the Class D amplifier, they will simply have to stop by for a visit.
Again, let me state one more time simply using a Class D amplifier does not mean your system will sound worse than a Class AB or Class A or even a tube amplifier. It does mean that if you don't have a noise problem then the design engineer who designed it dealt with the noise problem. Whoever was the design engineer on my Class D amplifier did not do enough homework or I would not have had the interference problem. |
Kijanki: Actually, a Class D amplifier uses a digital technique to reproduce the sound. The amp is actually a Pulse Width Modulator with a power output. Envision a one bit DAC reproducing the sound and you have an idea of what is going on in a Class D. Technically, it is incorrect to say it is a digital amplifier, the difference is described in the papers below. But do understand a Class D amplifier is certainly not a linear amplifier!
See John Guy's paper at http://www.national.com/vcm/NSC_Content/Files/en_US/Audio/ClassDAmplifierFAQ.pdf John is a very knowledgeable fellow, I worked with him at Maxim. Believe me when I say he knows Class D amps! The paper pretty well describes the Class D noise on the outputs.
John also published another paper in Audio Design Line which describes how a Class D amplifier works. I think you can get this magazine emailed to you directly if you are interested - I am not sure if you have to be an engineer or not to get it. Anyway, check out the paper at http://www.audiodesignline.com/howto/177102531;jsessionid=CP1YPZLBM2M5AQSNDLRSKH0CJUNN2JVN This is one of the best papers I have seen on how a Class D amplifier works.
It helps to know some algebra when you read these papers, as well as know a little bit about basic electronics. Otherwise, they are a good read.
It is incorrect to say I don't like my Class D amplifier; actually I have no plans to replace it. In fact, scroll up a ways and note what I said about listening to it with a CD as a source last week! That sound and bass was excellent! What I don't like is the radiated noise from the unit.
Let me explain the difference between conducted and radiated. Conducted noise is conducted along the wires, say the line cord, RCA cables, or speaker wires. Radiated means the noise is traveling through the air, much like a radio broadcast. If the noise is conducted, it will disappear when you unplug the offending wire or at least diminish when you put some ferrite suppression on it. If the noise is radiated, then simply moving the affected gear farther away from the noise source will reduce or eliminate it since radiated energy decreases as a square of the distance.
Note that sometimes it is difficult to determine if it is radiated or conducted. For example, if the noise was conducted from the AC power cord, unplugging it would also kill the unit. The only way to find out if it is conducted from the power cord in this case is to swap out power cords. It is further complicated by the fact the noise can be conducted out of the chassis by a cable and then radiated from there.
The FCC Class A you mention above is required for everything that is sold which has a clock in it above 10 KHz, I believe. The Class A or Class B spec dictates how much noise is allowed to pass, per FCC. Class A is allows ~10dB more noise in selected frequencies and is normally used for industrial equipment. Class B is normally used for consumer goods, such as radios, TVs, or stereos with computer chips in them. Class B is significantly more strict with respect to radiated and conducted noise. Actually, I am a bit surprised to see audiophile grade gear sold with a Class A level, but the intended application is what normally dictates the A or B designation, not who buys it.
Also, just because the frequency goes up doesn't mean the harmonics diminish. Once you are above 1 to 10 MHz or so, the phase-coherent relationship outside the chassis dictate how significant the emissions are. I have worked in EMI suppression before and it is a common mistake to assume a 2 MHz clock won't have significant or fail Class B emissions at 800 MHz, but don't count on it. I have seen units fail at those frequencies with a clock in the 2 MHz region. FCC requires testing up to 1GHz and sometimes higher just for this reason.
Also note when I say bandwidth, I don't mean the audible bandwidth at 20 KHz, I mean the bandwidth of the amplifier. If a given amp has a bandwidth of 60 KHz, then any injected noise in the region up to 90 KHz or so will have an effect on the sound, depending on how fast the response rolls off. But, lets keep this in perspective - one can inject a tiny amount of 60 KHz noise and the damage is causes may not be audible to our ears depending on our ability to hear it and our systems to resolve it, even though the sound did change a bit in the audible octaves.
I can give you a splendid example of this in one of my own designs. About two decades ago I was consulting for a successful audio amplifier company. They had a real winner of a power amp, it was listed in all the recommended components. Their preamp left a lot to be desired though, and that is why I was hired. I sketched out a phono preamp over dinner one night with the CEO of the company and their primary engineer.
I built it a few days later and plugged it into my system. The line stage was about perfect but the phono stage was a bit on the hard sounding side. Try as I might over the next week, I could not find the reason for the hardness. Since we were having a design review on Saturday, I had to fix it and fix it fast. Finally, around midnight on Friday, I found the problem. There was a 6 dB bump in the frequency response at 14 MHz and that is what was causing the hardness. Once I found the source for that, the hardness was gone completely. That bump at 14 MHz caused audible effects in the 1 to 3 KHz range even though at 14 MHz the gain had already rolled off quite a bit, beyond the -6 dB point. The gain was still significantly higher than unity though and that was the problem. Like I said, if it is in the bandwidth of the amp and the noise is bad enough, you will hear it! |
Kijanki:
I did go to the Icepower site and I downloaded his thesis. I skimmed through it yesterday and today at lunch and I focused on some of his major points. There is a lot there and he is certainly the first to put phased pulse code modulation into a Class D amp. I can see the advantages of that since it allows one to effectively have a higher sampling frequency while using a lower clock frequency.
But he did not invent phased pulse code modulation! I am pretty certain I worked on a PPCM circuit when I was in the Air Force working on some old DEW radar equipment; I know it has been used in various circuits since then as well.
You asked me to explain distortion mechanisms in Class D amplifiers. Either you did not read the links I posted above or you did not understand them as the biggest distortion mechanisms for Class D are described there. However, given that those papers were written for electrical engineers and not the layman, I think I will describe those in a separate post later. I will also include Class A and Class AB distortion mechanisms since they also apply to a Class D amplifier.
You are correct that Class D is analog, however it is most certainly not linear. Class D amps have much more in common with digital amplifiers or switching power supplies than linear Class AB amps. If I used the word analog in place of linear somewhere I apologize as that would be the wrong term.
However you are absolutely wrong when you think that Class D amplifiers have no resolution limits. Indeed they have at least as much or more than Class AB. The reason for this is that in a Class D amplifier there is a quantization error that exists because the modulator is trying to approximate the area under the analog input waveform in a given time sample using a varying pulse width approach, that is a PWM Rectangular waveform. As soon as you use that approach you develop a quantization error which does not exist in linear Class A or AB amps unless the units are severely bandwidth limited and even then it would not be a quantization error, it would simply be rolled off. I dont know who told you or why you would think a Class D amplifier has infinite resolution but it most certainly does not. There is nothing available at any price, using any form of amplification, such as tube or solid state, in an amplifier which has infinite resolution.
Using the IcePower approach with phased PWM one can obtain a reduction of a factor of four in quantization error, but that does not mean some number X divided by 4 is zero. Hardly! Note that if the sampling frequency was unscaled by four, you would have exactly the same thing as a phased PWM at 1/4 the sampling rate. The big advantage using the IcePower approach is for any given sampling frequency, the effects are equivalent to increasing it by a factor of four, or some integer multiple of that given the number of phased arrays you have.
But there is a trade off here: you cant increase the number of phased array output devices without dealing with increased problems in dead zone, turn on, turn off, rise time, and fall time problems. The more devices you put in parallel the greater problem you have with dead zone and the resultant harmonic distortion that results from it. The paper does mention this.
You also mention that 16 bit resolution with a 20KHz bandwidth is not possible unless you are sampling at 1.4 GHZ? Where did you get these numbers? I have news for you: we have an honest 16 bit systems at 196KHz and 20 KHz bandwidth working in the world today and in consumer gear, too. The big limitation in achieving 16+ bits of resolution is NOT in the digital domain but in the analog. There are few linear or analog ICs available that produce an honest analog equivalent in Signal to Noise ratio of more than 18 bits. Maxim has an audio IC which is guaranteed to produce 16 bits SNR, bu the price of that is so expensive hardly anyone uses it.
Discreet 16 bit SNR circuits are possible but add other issues In short, we are a long ways from true 24 bit SNR on the analog side. The clock speed is just not a problem and neither are the digital chips to run at that speed. Also, if you are speaking strictly in the domain of Class D, there is a LOT more to deal with than just the clock speed. Besides as stated in the thesis, it is clear that an effective 1.6GHz sampling rate is reasonable in production.
I disagree with your statement that only even harmonics are possible in Class D amplification. This is nuts really; if you look at the papers I posted above you can see that a dead zone in the region of 20 MHz to 500 MHz bandwidth causes as much as 2% harmonic distortion and this is not just even harmonics. And we did not even discuss clock jitter, whose effects are amplified by phase shifted pulse code modulation. Besides, the claim that odd harmonics are more audible than even harmonics was dispelled at least a decade ago. Music itself produces both even and odd harmonics as do loudspeakers. The Class D amplifier is very much like a push pull output stage in a linear amplifier, both even and odd harmonics will be produced. Even the thesis published distortion curves which are virtually identical to Class AB amplifiers.
Lastly, you continue to quote to me reviews by noted authors in various stereo review magazines as if that is supposed to prove to me that Model X amplifier is obviously a fantastic product. Frankly, if someone publishes something good about some product, I consider it is worth a listen but nothing beyond that. I hardly think for an instant that just because someone published an article stating that Model X amplifier it as close to perfect as possible and it is indeed so worthy of my cash that I will plunk down money for it. It might be great for their ears but I wear mine, not theirs. Consequently, I listen critically before I spend money.
I quit listening to that sort of review blabber at least 20 years ago when I walked into an expensive stereo salon in Beverly Hills, CA, and the sound there was substantially worse than what I had in my living room. And, the salespeople there were trying to tell me what they had was as perfect as I could possibly buy!
I walked out and didnt bother to spend a dime there, considering when I walked in I was ready to dump, in 1981 dollars, around 600 dollars on a phono cartridge.
You stated that listening is proof of concept, I do agree with that statement but did you read above where I did listen to my amplifier and I came to the conclusion after much listening over a period of time that my bass is somewhat dead and lifeless? I didnt measure this, I didnt read it, I determined that by listening to it. Actually, the Class D amplifier is only amplifier I have yet to put on a bench and test.
Perhaps your bass is fine, perhaps not, I dont know but I can assure you, my bass is not. Frankly, the bass was somewhat better when I had the Amber 70s in the system; at least until they ran out of voltage. It is hard to compare the sound of a 70 Watt amplifier to a 450 Watt amplifier when the 70 Watt amp clips occasionally. To my knowledge, my Marantz Class D amp has not clipped once. But that doesnt change the fact the bass now has a problem.
Kijanki, if you really believe that your Class D amplifier has 140 dB SNR, then do this experiment: take a battery powered AM radio and tune it to a spot on the dial which has no station. Then move it close to your Class D amplifier and see if it picks up any radiated EMI. An amplifier that has 140 dB SNR will not make a sound on the radio.
Frankly, I believe if you got that 140 dB SNR figure from some literature, they sure arent measuring it with respect to 1 Volt! Not to mention the fact that to measure the output of a Class D amplifier, one has to use a very steep filter to squash the noise on the output. Gosh, if I put a 5th order filter cut to 20 KHz on the output of a Class AB amplifier, I would get fantastic SNR too!
My advice to you is to start listening to your ears, accept the fact you can't afford the most expensive gear, and quite accepting reviews as gospel. In the long run, you will have a better sounding stereo and more money in your pocket. |
Kijanki, I sure don't know where you get your ideas on distortion but Class D amplifiers are not immune to any of the distortions that also affect Class AB or Class A. Unfortunately, using a Class D approach does not design any of them out. Shortchange a little output current during a dynamic burst of music and you will hear a lot of TIM, Class D or not.
Your analysis of 20 KHz, 16 bit resolution, and Class D is wrong. You are confused between the number of bits and the sampling frequency. The number of bits, which is exactly the same thing as the number of possible codes, dictates the maximum dynamic range available in the system while the sampling frequency dictates the maximum bandwidth. The analog equivalents of this is amplifier bandwidth equates to sampling frequency and the maximum voltage out equates to the number of bits in a system. How large in voltage each of those bits represent is dictated by the magnitude of the LSB. The entire reason for using a high frequency clock in the PWM is to include all the bandwidth of audio and reduce the quantization error.
Class D amplifiers can achieve the equivalent of 16 bit resolution if you can reduce the noise level to below 98 dBv on the output within the entire bandwidth of the amplifier. I sure don't see any reason why any well designed Class D amplifier can't do this. That of course doesn't mean they all do it but it certainly is very realistic and possible. This signal to noise ratio means it is measured on the output without the benefit of any external measurement filters. Unfortunately, most of the time Class D amplifiers are not measured in this manner so only the deign engineer knows what it is really capable of producing..
Here is another way to look at it: The human ear has the equivalent dynamic range of around 21 to 22 bits. Within that dynamic range, it can slide a window of 60 dB or approximately 8 to 10 bits or resolution. If a Class D amplifier was incapable of producing the equivalent of 16 bit performance then it is incapable of EVER achieving what the ear can hear. This is simply untrue.
That is a pretty amazing result regarding your little am radio. How close did you have it? Are we talking less than an inch or several feet? Did you try moving it in all three dimensions? Did you have the amplifier playing loud music at the time? Did you try setting the radio next to the speaker cables? If you did all this and you didnt hear a peep on the radio then I would sure love to see that!
Measuring the signal to noise ratio, or SNR, within a bandpass of 20 Hz to 20 KHz means the signal to noise figure is realistic to what noise you can hear; it is not realistic to what damage the noise can do to the music. Measuring SNR wideband can indicate how much noise intermodulation distortion an amplifier can have and you will hear the effects of that even though the SNR within the audio bandwidth is very good. I guarantee that the 140 db SNR you quoted above was measured with a measurement filter in front of the analyzer. Where did you see this spec? Is it published somewhere? I like to see under what conditions they measured it. I suspect it is measured with an aggressive filter and is probably relative to something other than 1 V out.
Just out of curiosity, what model Class D amplifiers to you have? And what did they cost?
Guidocorona: Your Socrates logic is missing a few logic equations. You can say that and you can say you used logic to achieve it but you would be wrong.
I never said Class D couldnt sound good or even sound great. All I have been trying to do is explain to some folks why Class D amplifiers have a steeper curve to achieve good sound than Class AB or Class A. But it seems that anyone who owns a Class D amplifier is oblivious to the concept that their amplifier can produce, heaven forbid, unmusical distortions or any distortions that Class AB amplifiers also produce. Furthermore, it seems on this board that anyone who speaks in the least unfavorable manner of Class D is automatically considered hysterical. I have to wonder if these folks are in emotional denial or something.
Short of winning the lottery, I have no real interest in purchasing another amplifier. For now and for the next few years, this Class D amplifier I have will have to suit the purpose of providing power for the subwoofers. When it comes time to upgrade it, I will try a few amplifiers but in my system, not listening to it in some unknown stereo I have never heard before.
I would try a few more audio stores but how can one audition a given amplifier when the entire system is new to my ears? Is what I hear the amplifier or something else? Is the system masking some of the distortion mechanisms in the amplifier so it sounds great? This is why when I walk into an audio store, it is more to discuss what is new in the audio world, see what the store has available, and perhaps gather a few opinions on what I should audition. If the music is playing and it sounds great then I get a bonus, too. Unfortunately, about half the time the sound is worse than I have at home and that can be pretty depressing given I am in the market to upgrade.
Remember, I am a design engineer, so I look at things differently than most audiophiles. I am far more interested in who the designer is, what his/her opinion is on how to achieve good sound, what they did to get there, and how they got there. I try to find out where they put their design emphasis. When I meet someone who is blowing smoke in my face from an out of control ego and trying to sell me something which is obviously a lie, I know none of his/her products are worthy of my listening time. Frankly, I rarely let them know I have a degree in engineering much less that I actually design audio gear. I can learn so much more just playing dumb and letting them talk. There are a lot of honest design engineers out there, please dont get me wrong. Actually, I believe the proper term is most; they are quite honest and deliver what they believe to be an honest product. But there are a few whoppers out there, that is for sure.
Also, to say there is no single test that describes the sound of an amplifier is correct. But it is incorrect to say there are no tests that tell you anything about the sound of an amplifier is completely wrong. It seems that there is a belief that measurements do not tell you how an amplifier can sound but that is not entirely true. One can learn a lot about how an amplifier will sound before you even hook up speakers with nothing more than a good audio analyzer. It is only a matter of knowing where to look, what to test, and how to test it.
Muralman1: I am intrigued by your comments. When you say you non oversampling DACs are only appropriate do you mean that oversampling DACs actually make the sound worse? That is, you have an oversampling DAC in front of the Class D power amplifier and the sound is worse than if you used a DAC which did not oversample? If so, I find that very interesting. Also, what do you mean by a full spectrum DAC? Spectrum usually implies bandwidth related characteristic -was the bandwidth changed in someway? |
Muralman1: An interesting mantra that you have. Consider this along that line of thought: "If our understanding of Class D is deficient, which is in deed a valid argument, then also too are the designers whose understanding is also deficient, which would then yield a deficient amplifier." So, it works both ways.
There is no question in my mind that Class D will get better with time. Better devices and better deign techniques. However, if the bulk of Class D end up in low to mid fi battery operated speakers, then it will be a very long learning curve indeed. As one Class D IC designer told me a while back, efficiency is the sole purpose for Class D's existence, otherwise there is no point. I don't totally agree with that statement, as it is clear that Class D allows for great bang for the buck in higher power amplifiers. I am sure my Marantz amplifier would have cost at least 5x more for the same given output power and the same given number of channels.
I agree entirely that the cable industry is a house of cards. Although I can see the reason for the expense in a number of expensive cables, I do have to question the reasons for the existence of some of the expensive cables indeed. Granted test data does not tell all, but the lack of any test data of any kind is especially suspect to me in the more expensive cables. This is especially true given the great cable test equipment we have today.
Guidocorona: I will check out the review and look for an amplifier here locally. I am curious indeed. I think I might have to pack a small AM radio, too. |
