Switching Amplifiers: The Technology and the Issues - page 2
Feedback from Industry Professionals
So is there any problem here at all? The first and foremost problem I see here is consumer audio has no standard for grounding, and these out of band artifacts from the switching amps could well "pollute" the grounds of any equipment connected to the unit. For any digital input this may not be critical, for any analog inputs the effect could well be much more severe. However Mike Danielson has said the following " Your comments on lack of grounding standards in consumer audio and polluting other systems are dead on, but the signal will not fold back into the audio band as noise. The filters that raise the sampling frequency 8x to the output are carefully designed to avoid this and are used specifically to shift any audio band noise up above the audio band, this is exactly how DAC's work. Test equipment how ever does not respond gracefully to this noise."
This is true, but we should note that DAC's usually come with a low pass output filter to attenuate those artifacts by 40db or more.
Further Comments on grounding and PCB layout from Bruno Putzeys:
"A practical problem that pops up is that we can no longer use the chassis as a faraday shield in applications that have large HF signals (such as class D amps) or high impedance circuits (volume control). Instead, local shielding connected to audio ground should be used.
The difficulty lies in audio folks' startling inability to make good circuit board layouts. If the use of grounding among the boxes is mildly muddy, grounding on a circuit board level is a total disaster. You will often find bits and pieces of ground plane flying around, signal ground traces, power ground traces haphazardly interconnected ("wow! Lots of thinking must have gone in this ground layout! The designer must be a real expert!"). As if electrons follow road signs saying power return this way or signal return that way. Equally common, all four edges of the circuit board are used for placing connectors on. Apparently to insure that all ground currents from one connector really do travel through all the nice audio circuits before getting to the other connector.
If an RF current is injected on the shield of an audio cable, this current will run around the whole chain, creating voltage drops. These voltages will get demodulated inside the circuit. So, people then start adding small caps from audio ground to chassis at the RCA inputs. (Another RF ground loop! The boss will be happy). This is of course where the misery starts, because now the chassis (and hence the earth ground) is back in the equation. "
Bruno continues: "The difficulty lies in the fact that life isn't getting easier in the future. Switch-mode power supplies suffer an aggravated version of interwinding coupling across the transformer. The circulating current between the primary and secondary sides is quenched by placing a so-called "Y" capacitor between primary "ground" and secondary ground. This reduces the voltage imposed by the transformer between the two sides by a factor equal to the ratio between Y capacitance and parasitic capacitance between the windings. Increasing the Y cap reduces the problem, but never solves it. Worse still, the Y cap effectively connects the AC line to the audio ground! Similarly, other EMI problems are "fixed" by connecting caps to chassis everywhere. Y caps are also found sometimes in audio equipment that has a linear supply and a good reputation. They are also part of mains filters, devices often used as a selling argument in the audio world (cleaner mains! sure! but the dirt is now in your audio).
Such practices are of course of no consequence in computers and copiers, VCRs and TVs (when not connected to the audio system). Unfortunately the same people who have been smart enough to design these are now selling power supplies for use in DVD players etc.
Designing a Switch-mode power supply that does not create circulating currents and that does not need Y caps is certainly possible. I am afraid, however, that the number of designers worldwide capable of pulling this off can be counted on one hand, and they might not even actually be at it."
Over the years in designing, and manufacturing linear design audio equipment I have come across some rather dubious approaches (or lack there of) to grounding techniques in most consumer unbalanced gear. Bruno's observations are just dead on. Since consumer audio designers have never felt required to rigorously address these issues for low frequency linear design, FM tuner design, and CD player design, I have real doubts as to how well this will be addressed in switching amplifier and switching power supply design. For more commentary on this please refer to page two of the article I wrote on the Yamaha T-80 Tuner . Note the picture of the added ground straps and the installation of isolated RCA input jacks.
The grounding, shielding and board layout issues have been a hot topic on one consumer board in particular. The following links are a perfect example of the problems consumer audio has manifested upon itself.
High Frequency Radiation Issue
The second issue is the HF components that are going out to the speaker. These artifacts in themselves may well be of little importance, but I have observed the following additional ultra-sonic sources in my own lab:the Ionic Breeze and the Environizer, both air cleaners. The Ionic Breeze unit will emit triangle waves at about 16 kHz and the Environizer unit will emit 66 kHz bursts.
I discovered this when I was testing an open-chassis high-gain board on the bench, which picked this up rather nicely, so to speak. These sources are benign individually,, but the combination of a number of different sources of ultrasonic energy could well induce an interference issue for consumer audio gear. I would not consider this to be an issue if consumer audio gear was well grounded and shielded, but since it isn't, I do have concerns.
Do all switching amps have the same output spectrum as the Panasonic unit? The Panasonic unit uses a binary switching approach, and DDX technology uses another approach. Note the following from DDX.
Filtered PWM output: Note that DDX tristate modulation provides significantly lower ripple with the same output filter.
(PWM switching ripple on Audio output (DDX is a registered trademark of SigmaTel Inc.)
Please note the DDX modulation scheme gives a 16 db decrease in output ripple for a simple LC filter as shown below.
Typical output filter - 2nd order LC Butterworth tuned
Common values: L1, L2 10-22uH
C1, C5 0.1-0.22uF, C2, C4 0.1uF
R1, R2 5-7 ohms, C3, 0.47-1.0uF
C6, C7 1000pF, Zload 4-8ohms
R1, C2 and R2, C4 provide damping of the L1, C1 and L2, C5 filter response
C3 is the differential cap that does the majority of the work from the audio perspective
C6, C7 provide some addition filtering for high frequency radiated emissions above 30MHz
Damping Factor/Output Impedance and Switching Amplifiers
Damping Factor, which is really a misnomer and should only be named output impedance, has seen quite a bit of controversy in audio. Use of the term damping factor in this article will refer to the ratio of output impedance of the amplifier to an 8 ohm resistive load. There is no connection implied that high or low output impedance will effect mechanical damping of the loudspeaker.
Linear power amplifiers typically have a small choke (2 to 10 micro henrys, typical) after the feedback loop to prevent oscillation into difficult capacitive loads. Practically all linear power amps have a Zobel output filter (a resistor and capacitor in series from the output to ground) for stability. So the output impedance of linear amplifiers was generally rather benign. If memory serves me correctly, DBX back in the 1980's found that in some cases the small choke at the output of the power amplifier did interact with the loudspeaker crossover and caused frequency response errors and in some cases distortion. The distortion appears to be primarily a function of the amplifier in front of the load and how well it is compensated for reactive load stability. This finding is not surprising because the ideal audio amplifier should have as low a reactive output impedance as possible when driving a complex reactive load such as a loudspeaker. Here at R. E Designs, I've found that it certainly is possible to build a linear power amp without an output choke and still maintain good stability into complex reactive loads.
The question still remains is just how high or low should the output impedance be?
Again, previous papers and articles have suggested the very minimum should be a damping factor ( Note: the term damping factor is really a misnomer because the output impedance of the amplifier has little or nothing to do with how well a speaker is internally damped. The correct term should be output impedance !) of twenty, or 0.4 ohms for an 8 ohm load. Many of us decided that this was not enough considering the impedance swings in loudspeakers and decided that if the output impedance was below 0.1 ohm than this would accommodate 99% of the loudspeakers on the market. In addition, the low output impedance would keep this out of the equation so to speak. Personal preference also plays a role here. Some audiophiles prefer the higher output impedance for that "tube like" sound while some audiophiles don't.
Most of the switching amplifiers on the market today have the LC output filter after the feedback loop. By definition we now have a complex output impedance driving a complex input impedance, and since loudspeaker crossovers have plenty of variation, frequency response issues and added distortion can be a factor. It is clear that most of the switching amps out there today really need a resistive load to perform at their optimum.